Beyond Ses: A Resource Model of Political Participation

Author(s): Henry E. Brady, Sidney Verba and Kay Lehman Schlozman

Source: The American Political Science Review, Vol. 89, No. 2 (Jun., 1995), pp. 271-294
Published by: American Political Science Association
Stable URL: http://www.jstor.org/stable/2082425
Accessed: 18-01-2018 21:27 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

American Political Science Association is collaborating with JSTOR to digitize, preserve and
extend access to The American Political Science Review

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
American Political Science Review Vol. 89, No. 2 June 1995

BEYOND SES: A RESOURCE MODEL OF POLITICAL PARTICIPATION

HENRY E. BRADY University of California, Berkeley
SIDNEY VERBA Harvard University
KAY LEHMAN SCHLOZMAN Boston College

is paper develops a resource model of political participation. The resources considered are
time, money, and civic skills-those communications and organizational capacities that are
essential to political activity. These skills are not only acquired early in life but developed in
the nonpolitical institutional settings of adult life: the workplace, organizations, and churches and
synagogues. These resources are distributed differentially among groups defined by socioeconomic
status. A two-stage least squares analysis shows these resources have powerful effects on overall
political activity, thus explaining why socioeconomic status has traditionally been so powerful in
predicting participation. We disaggregate overall activity into three kinds of acts: those that involve
giving time, those that entail donating money, and voting. Each requires a different configuration of
resources resulting in different patterns of stratification across various political acts.

W hy do citizens participate in political life? show how the importance of a resource depends
One way to think about this puzzle is to upon the particular activity. Education, for example,
invert the question and ask why people is important for some political activities because it
don't take part in politics. Three answers immedi- enhances political interest and civic skills while in-
ately suggest themselves: because they can't, because come is important for other activities because of the
they don't want to, or because nobody asked.' "They monetary resources it provides.5 By showing how
can't" suggests a paucity of necessary resources: time resources differentially available on the basis of SES
to take part in political activity, money to make con- affect various modes of political activity, we explain
tributions, and civic skills (i.e., the communications not only why some individuals are more active and
and organizational skills that facilitate effective par- others less but also why certain kinds of people
ticipation). "They don't want to" focuses on the engage in particular kinds of political activity.
absence of psychological engagement with politics-a A resource-based approach also has methodologi-
lack of interest in politics, minimal concern with cal and theoretical advantages, especially in compar-
public issues, a sense that activity makes no differ- ison to explanations based solely on psychological
ence, and no consciousness of membership in a engagement with politics, thus yielding a more pow-
group with shared political interests. "Nobody erful explanation of participation. We are more con-
asked" implies isolation from the recruitment networks fident in our ability to measure resources than in our
through which citizens are mobilized to politics.2 ability to measure psychological engagement. Re-
All three factors help explain political participation, ports of attitudes are notoriously fugitive, unreliable,
but we focus on the role of resources-time, money, and difficult to compare across respondents (Duncan
and civic skills3-for explaining political participation 1984). As we shall see, the measurement of resources
in America. Adding resources to the other two expla- rests on more factual questions for which the metrics
nations permits us to move beyond the "SES model," used-dollars, hours, and the number of letters writ-
that is, beyond explanations of political activity based ten or speeches given-are unlikely to vary in mean-
on one or more of the components of socioeconomic ing from respondent to respondent. Although re-
status: education, income, and occupation.4 By at- sponses may suffer from the fallibility of human
tending to resources conceived at a general level, we memory, at least these questions are about concrete,
can probe the way resources link backward to SES everyday matters. Furthermore, when linking en-
and other social characteristics and forward to polit- gagement and activity, it is hard to be certain of the
ical activity. Going backward from resources, we can direction of the causal arrow. Political interest and
show that the three resources of money, time, and political efficacy, for example, certainly facilitate po-
civic skills vary in their association with SES and litical activity, but activity presumably enhances in-
other social characteristics. Money and some kinds of terest and efficacy as well. Indeed, most measures of
civic skills are closely related to SES, but time and psychological engagement with politics are, by their
other civic skills are less stratified. Civic skills are less very nature, perilously close to activity itself. This
stratified by SES partly because social characteristics makes them robust predictors of political participa-
such as affiliations with "congregational" churches tion but trivial (and possibly spurious) explanations
are not highly correlated with SES and these affilia- for participation. Yet despite its apparent "head
tions serve as training grounds for civic skills. As we start" as a predictor of participation, we show that
go forward from resources to political activity we can political interest-a standard measure of psychologi-

271

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 A Resource Model of
cal engagement in politics-does not displace re-
sources as a predictor of political participation.
Theoretical considerations lend additional support
to the notion that a resource model can provide a
powerful explanation of political participation. Unlike
psychological engagement in politics, which probably
develops along with political activity, the institutional
involvements from which citizens acquire resources
generally antedate or occur independently of political
activity. Obviously, family background and early
experiences in school-critical for both the develop-
ment of resources for politics and for the future
institutional commitments that permit the further
enhancement of political resources-precede adult
activity. In the absence of actual life histories col-
lected over respondents' lifetimes, we cannot be
absolutely certain that adult decisions about family,
work, organizational involvements, or affiliations
with religious institutions are apart from and in
advance of choices to take part politically, but these
seem plausible assumptions.6 In addition, a resource
theory has implications for the normative issue of
how we construe political inactivity, especially when
the politically quiescent have obvious and pressing
needs. If individuals eschew politics because they do
not care-because they prefer to devote themselves
to private rather than public pursuits-then we are
apt to dismiss inactivity as a matter of personal
choice. If the failure to get involved is the conse-
quence of resource constraints that make it difficult
for even those who are politically interested and
engaged to take part in political life, then we are
likely to be more concerned about political inactivity.
Finally, resource models of political participation
tie into two powerful intellectual traditions: stratifica-
tion theories from sociology and individual choice
perspectives from economics. The SES model follows THE CITIZEN PARTICIPATION STUDY
naturally from stratification theories,7 which suggest
that class and status hierarchies are fundamental
features of modem industrial societies that often
determine their politics. True to the theory, the SES
model does an excellent job predicting political partic-
ipation.8 Yet the SES model fails to specify clearly the
mechanism linking social statuses to activity. Rational
choice theories, on the other hand, have clearly
specified how and why individuals might decide to
participate in politics to pursue their self-interest, but much longer, in-person interviews were conducted
these theories have done a very poor job predicting
political participation.9 Indeed, with respect to SES
and participation, at least one variant of the rational
choice approach suggests no relationship at all or, if
any, that people of high SES (who by virtue of their
high levels of education command the intellectual
sophistication to comprehend the free-rider problem
and who by virtue of their high salaries would find
the opportunity cost of participation prohibitive) to
be least likely to take part in politics. Instead those
with high levels of SES, who are not otherwise
known for particular irrationality in the conduct of
their lives, are the most likely to be active.10
The problem is that rational choice approaches
have focused on how the benefits of participation
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
Participation June 1995
might offset the costs of participation without exam-
ining costs very carefully. A resource perspective
takes seriously the costs of using resources. In doing
so, we apply to politics an important variant of
rational choice theory. The Chicago school of eco-
nomics" has shown how a powerful theory of choice
can be built not upon restrictions on the motives for
choice (as the emphasis on self-interest over altruism
or duty does) but upon the budget constraints on
resources that limit choices. If there are multiple
constraints on a series of resources that vary indepen-
dently in the population, then a formidable theory
can be based upon the degree to which each resource
constraint is binding in a particular situation. And
since resources such as money, time, and skills can be
measured and affected by policymakers, it is useful to
formulate a theory based on a careful description of
how variations in resources flowing from social strat-
ification enable and restrict individual activity.
To develop a resource model of political participa-
tion requires four steps. First, we define resources
and explicate how we measure them. Second, we
show how resources are distributed in the popula-
tion, in particular, how they relate to SES. Third, we
look closely at the resource of civic skills (particularly
those acquired as an adult) in order to show that our
somewhat indirect indicators of civic skills indeed
measure a single dimension of civic skill that is
developed in extrapolitical institutional settings and
available for political activity. The fourth step is the
heart of our enterprise-we show that resources
explain political participation and that different re-
sources are related to different activities.
Our data come from a large-scale, two-stage survey of
the voluntary activity of the American public. The
first stage consisted of over 15,000 telephone inter-
views of a random sample of the American public
conducted during the last six months of 1989. These
20-minute screener interviews provided a profile of
political and nonpolitical activity as well as basic
demographic information. In the spring of 1990,
with a subset of 2,517 of the original 15,000 respon-
dents chosen so as to produce a disproportionate
number of both activists as well as African-Americans
and Latinos. In the following analyses, we have
reweighted the follow-up sample so that we have
a representative sample. (See Verba, Schlozman,
Brady, and Nie 1993 for a description of the sample.)
The study is unusual in focusing on voluntary
activity not only in politics but also in churches and
organizations. In addition, we construed political
participation quite broadly, including not only voting
and other forms of electoral activity (e.g., working in
campaigns, making financial contributions) but also
contacting public officials, attending protests, and
getting involved either formally or informally on local
272
 American Political Science
issues. (Appendix A provides a list of the questions
used in this paper.)
DEFINING AND MEASURING
POLITICAL RESOURCES
We begin by outlining the resources for political
participation.
Time and Money
The two prime resources for investment in political
participation are time and money. Individuals use
time in the service of political action in many ways
(e.g., working in a campaign, writing a letter to a
public official, attending a community meeting).
Money, of course, can be donated to candidates,
parties, or innumerable political organizations or
causes. We measure money resources by family in-
come from all sources in $10,000 units.12 We
measure
the "free" time available for political activity by the
hours, if any, left over after accounting for time spent
in an average day doing work for pay, doing neces-
sary household work of all sorts, studying or going to
school, and sleeping.'3 Time and money differ in
significant ways as resources. In comparison with
money, time is both more constrained and more
equally distributed-everyone has only 24 hours in a
day. The upper limit on money, of course, is much
less constrained, and differences among individuals
can be much larger. Time is constrained in another
way that affects the way it is distributed. Time not
used today cannot be put in the bank. Money, in
contrast, can be accumulated for later use.14
Civic Skills
Civic skills-those communications and organiza-
tional capacities that are so essential to political
activity-constitute a third resource for participa-
tion. 5 Citizens who can speak or write well or who
are comfortable organizing and taking part in meet-
ings are likely to be more effective when they get
involved in politics. The acquisition of civic skills
begins early in life-at home and, especially, in
school. However, the process need not cease with the
end of schooling but can continue throughout adult-
hood. Adult civic skills relevant for politics can be
acquired and honed in the nonpolitical institutions
of adult life-the workplace, voluntary associations,
and churches. Managing a reception for new employ-
ees and addressing them about company benefits
policy, coordinating the volunteers for the Heart
Fund drive, or arranging the details for a tour by the
church children's choir-all these undertakings rep-
resent opportunities in nonpolitical settings to learn,
maintain, or improve civic skills.
We measured civic skills in several ways. Since
communications and organizational skills are ac-
quired in school, we asked an extensive set of ques-
tions about educational attainment.'6 From these we
273
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
Review Vol. 89, No. 2
constructed an eight-point scale ranging from a gram-
mar school education to a Ph.D. or professional
degree."7 In addition, one particular aspect of ed
tional experience-participation in student govern-
ment in high school- provides a potentially useful
measure of civic skills.'B We measure this by a four-
point scale ranging from no activity to very active.
Since facility in expression is central to the ability to
communicate effectively, we also use two measures
of language ability, testing somewhat different no-
tions of how verbal facility matters. American society
puts a premium on speaking and understanding
English, so we asked our respondents what language
they ordinarily speak at home: English (scored as 3),
a combination of English and another language (2), or
another language (1). We thought that those who did
not speak English at home would be less likely to find
it easy to participate in politics. In addition, to assess
verbal ability, we used the score (number of items
correct) on a 10-item vocabulary test that has been
used regularly since 1974 on the National Opinion
Research Center's General Social Survey.19 For the
1%o of our sample interviewed in Spanish, this test
was administered using Spanish words.20 Alwin
notes that this vocabulary score is strongly related to
schooling (a correlation of .51 in our sample and .54
in Alwin's), but such scores also "correlate highly
with tests of general intelligence-usually .8 or high-
er-and are good indicators of scores on the verbal
component of standard tests of general intelligence"
(1991, 627).21 Years of education and vocabulary score
are included in our model to show that both school-
ing and general intelligence matter for political par-
ticipation and to provide additional support for our
contention that civic skills matter when other factors
are controlled.
To measure civic skills developed as an adult, we
asked those with jobs and those who reported activ-
ity in a church or an organization22 whether as part
of their involvement in each sphere, they had, in the
past six months, engaged in the following activities:
written a letter, gone to a meeting where decisions
were made, planned or chaired a meeting, or given a
presentation or speech. In each realm (on the job, at
church, or in nonpolitical voluntary organizations)
we measured civic skills as the number, ranging from
zero to four, of these skill-acts undertaken by the
respondent in the last six months. Those who have
an opportunity to do these things in a nonpolitical
setting would, presumably, be more willing and able
to do them in a political context. In this sense, we
expect that these competencies can be used as inde-
pendent variables to explain political participation.
Table 1 presents, both for all respondents and for
only those involved in each arena, the frequency with
which individuals engage in activities that we expect
to produce civic skills. People are most likely to
engage in skill-acts at work: 53% of the sample
practiced at least one skill on the job in the six months
before the survey. Yet 33% of the population engaged
in skill-acts in nonpolitical organizations and 20% in
churches or synagogues.
A Resource Model of Participation June 1995

Percent Reporting Various Activities in Nonpolitical Settings

IN NONPOLITICAL IN CHURCH OR
ON THE JOB ORGANIZATIONS SYNAGOGUE
% OF
% OF % OF ORGANIZA- % OF % OF
WHOLE % OF WHOLE TIONALLY WHOLE CHURCH
CIVIL SKILL SAMPLE WORKINGa SAMPLE INVOLVEDb SAMPLE MEMBERSC

Attend a meeting where

decisions are made 48 70 30 41 18 27
Plan such a meeting 24 36 14 19 9 13
Write a letter 40 58 15 16 6 9
Make a speech or presentation 28 40 15 18 10 15
At least one of the above 53 78 33 44 20 29
Source: Citizen Participation Study.
aWorking full- or part-time; with a job but not at work due to vacation, illness, etc.
bMember of or contributor to an organization that does not take stands on public issues.
'Member of or regular attender of services at a local church or synagogue.

THE DISTRIBUTION OF Figure 1 shows the very different relationship be-

POLITICAL RESOURCES
As we shall demonstrate, the presence or absence of
resources contributes substantially to individual dif-
ferences in participation. Resources are, in turn, not
equally distributed; some socioeconomic groups are
better endowed than others. This makes a resource
model useful not only for explaining individual dif-
ferences in political activity but for explaining differ- however, greater educational advantage is associated
ences in activity among politically significant social
groups, especially differences along SES lines. How-
ever, resources vary in the extent to which they are
stratified by SES, that is, in the extent to which they
are differentially available to those high on the SES
scale.
Money and Time
If you want to give money, you must have money;
if you want to contribute time, you must have some
free time. Income, of course, is one of the compo-
nents of socioeconomic status. By definition, it is
concentrated in the hands of the wealthy; we also
know that those with more education and higher-
status jobs command a disproportionate share of the
wealth. We know less about the distribution of free
time in relation to SES. On one hand, the rich might
have more free time because they can hire others to
do what most people have to do for themselves. On
the other, the rich might have less free time because
they accumulate their wealth by logging long hours at
work. These conjectures reflect the contradictory pre- contrast to money, which is of a piece with SES, the
dictions of economic theory, which holds both that an
income effect would produce more leisure for the rich
because they are able to purchase it and that a
substitution effect would produce less because their
higher wages raise the opportunity cost of free time
(Mincer 1962). In fact, neither conjecture is correct:
free time and SES are unrelated.
tween money and SES (as measured by educational
attainment) and time and SES. As expected, educa-
tion and family income are strongly related. There
is no such consistent pattern of stratification when it
comes to time. Those in the least well educated
group-a disproportionate number (51%) of whom
are retired, keeping house, or permanently dis-
abled-have on average more free time. Beyond this,
with neither more nor less free time. If we were to
consider another aspect of SES-position on the
occupational hierarchy-the contrast between money
and time would appear even more clearly. If working
respondents are stratified into job levels based on
how much formal education and on-the-job training
their jobs require, we find, not surprisingly, that
family income rises sharply with each step on the
occupational ladder. However, this pattern is not
repeated for free time: those in the least skilled jobs
have almost exactly the same number of hours free
time per day as do those in the highest level jobs.
What then affects the availability of free time? The
answer is simple. The factors that affect free time are
"life circumstances": having a job, especially a full-
time one; having a spouse with a job; and having
children at home, especially preschool children-all
diminish the amount of free time available. Those
working full-time report, on average, six fewer hours
free per day; a working spouse reduces free time by
about three hours; preschoolers at home reduce free
time by about three hours.23 This finding-that in
amount of free time available varies with life circum-
stances rather than with socioeconomic advantage-
has implications for American politics. To the extent
that citizen politics in America relies increasingly on
modes of activity that use money rather than time
as a resource, the edge enjoyed by the already-
advantaged is enhanced.

274

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 American Political Science
Educational Stratification of Income and Free Time
Income In $1 ,?OO's Free Time In Hours Per Day
$70 1
$60T6 -
Is s 8~~~~~~~~~~~~~~~~
$50-
$30 - Average Hours Per Day Free 4
$20 - Average Yearly Income ($1,000's)
$0
No H.S. Diploma H.S. Diploma Some College College Graduate Some Graduate
Level of Education
Civic Skills
Civic skills are, in general, more likely to be pos-
sessed by the socioeconomically advantaged. Those
with higher levels of education are more likely to
speak English at home, to have better vocabulary
skills, and to have taken part in high school govern-
ment. Civic skills acquired as an adult at work, in
organizations, and in church are also stratified by
education. The stratification of civic skills by educa-
tion, however, varies greatly across jobs, organiza-
tions, and religious institutions. Figure 2 shows the
mean number of skills exercised in each of the three
institutional settings for those at various levels of
education. (The pattern would be the same if the
measure of SES were income or occupational level.)
All respondents are included in the computation of
average skills whether they are involved in a partic-
ular setting or not (i.e., whether they have jobs, are
affiliated with an organization, or belong to a church
or not). Those who never finished high school re-
ceive few skill opportunities anywhere. Those with
at least a high school diploma have more opportuni-
ties to practice civic skills on the job than in organi-
zations or in church, presumably reflecting the fact
that most people spend more time working than en-
gaging in organizational or church activities. The net
result is that in providing opportunities to exercise
skills, workplaces discriminate most-and churches
least-on the basis of educational attainment.
The process by which these results are achieved
varies across these institutions. The opportunity to
practice civic skills in an institution requires both
involvement in the institution and a setting that pro-
275
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
Review Vol. 89, No. 2
I
vides the chance to practice some skills. The stratifica-
tion by education for job-based civic skills comes
primarily from differences in chances to practice skills
and not from differences in attachments to the labor
force. Those with higher education are only slightly
more likely to be working than the less well educated,
but among those with jobs, the better-educated are
much more likely to have chances to practice skills.
The stratification of skill opportunities in voluntary
associations is somewhat different. The advantage of
the educated in this respect comes from the fact that
those with high levels of educational attainment are
considerably more likely than those at lower levels
to be involved with an organization. Among the
involved, however, there is less difference among
educational groups in the practice of civic skills.
Finally, churches are most egalitarian in the civic skill
opportunities they afford, and they are egalitarian in
two ways: (1) there is no consistent relationship
between education and church membership: those
with the least education are as likely as those with the
most to attend church regularly; and (2) among those
who attend church, there is relatively little stratifica-
tion by education in terms of who makes a speech or
organizes a meeting.
The differences across the three institutions are
significant for the stratification of participation in
American politics. The workplace reinforces initial
socioeconomic advantage as the well-educated com-
pound their advantage by developing skills on the
job. Since the educated join more organizations,
voluntary associations also reinforce earlier advan-
tage. However, organizations offer those affiliated
 A Resource Model of Participation June 1995

rage Civic Skills Acquired on Job, in a Non-Political Organization, and in Church

Average Number of Skills

4

* Church Skills -W organization Skiiis A-Job Skiiis

3.5 +

2.5

1.5

0.5

No H.S. Diploma H.S. Diploma Some College College Graduate Some Graduate

Level of Education

Among all respondents at each educational level

with them chances to practice skills with relatively church-provides opportunities to acquire the re-
little regard for educational attainment. Finally, be- sources relevant to political activity. Civic skills are a
cause church attendance is not stratified by SES and central-and, we believe, innovative-component of
because, within the church or synagogue, education our model. They are also somewhat problematic. We
plays a smaller role in who is active, religious insti- begin our analysis.of the resource model with a closer
tutions are the most egalitarian in terms of civic look at civic skills.
training. Compare, for example, the skill-developing
opportunities in workplace versus church for a Catholic
with a professional or managerial job (53% of whom DEVELOPING CIVIC SKILLS:
graduated from college) and a Baptist with a clerical A LEARNING MODEL
or blue-collar job (only 3% of whom graduated from
college). The Catholic professional or manager prac- Our measures of civic skills--educational attainment,
tices an average of 2.89 skills on the job but only .22 participation in high school government, language
skills in church. Compared to the Catholic profes- ability, and reports on activities in adult institu-
sional or manager, the Baptist clerical or blue-collar tions-are a disparate and somewhat indirect set.
worker averages fewer skill acts (only 1.11) on the job Therefore, we need to demonstrate that these are all
but much more at church (.84). indeed measures of civic skills and that these skills
To summarize, we have considered several kinds are developed or perfected in the institutions we
of resources. These resources are distributed differ- describe. If these conditions hold, we will have a
entially across socioeconomic groups. If these re- powerful set of variables that can be used to explain
sources, acquired outside of politics, affect political political activity. The task is somewhat easier for civic
activity, we will have a potent explanation of the skills learned in the home and in school. It seems
origin of disparities in participation across social reasonable to suppose that education, language abil-
groups. If the various resources are differentially ities, and participation in high school government
useful for alternative political activities, our model constitute real measures of skills that can facilitate
will also explain why some forms of involvement are political participation. And we know that they are
more stratified than others. temporally prior to adult engagement in politics.
Figure 3 displays the resource model schematically. However, we wish to demonstrate that civic skills are
We shall not discuss it fully here but refer to it as we also developed in adulthood and that these skills
move through our argument. At this point, let us then foster political participation.
simply indicate that Figure 3 summarizes the way in There are three obstacles to making this case for
which involvement with institutions-first in school civic skills in general and for adult civic skills in
and later on the job, in organizations, and in particular. First, there is a measurement problem. We

276

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 American Political Science Review Vol. 89, No. 2

The Resource Model

Institutional Resources: Political

Institutional Involvements Skill Skil Act Skills Interest and Political
Types and Education Opportunities lMel Self-described Acts

Type TimeSelf-sOpportunities S

Occupation Sil

Active in Organizational Organizational A x PgiilAt

Organizations Opportunities _ Skill Acts ~ _C_ Voting,
Contacting,
Campaign Work,
Attendan at / Civic Skills vCampaign Mon
Attendance at ~~~~~~~~~~~~~~~~~~~~~~~~~~Informal
Church / / // Community,
Time in71 1 A/ / / J Boards,
Church Work ~~~~~~~~~~~~~~~~~~~~~~~~~Protest I
Opportunities Acts/
Catholic or <,/
Non-Catholic / / _ n wa

| Church >L - - / Tm

Vocabulary /___/__

* I z ~ ~ ~ ~ ~one-way| Hi ~ ~ ~ ~ ~ Political
Observed Unobse/ed -nCitizenship
Years in School Fml

High School Income

Government- __

,-~~~~~~~~~~~~~~~ Interest

'Key: L k Jcausation ' *

* ~~~~~~~~~~~~simultaneous
Observed Unobserved causation *
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

measure civic skills acquired before adulthood indi- there might be a spurious correlation problem. A third
rectly by asking about educational experiences and variable (e.g., a general taste for activity) may lead
language abilities. It is not certain that those with individuals both to practice skills in nonpolitical in-
high levels of educational attainment, experience in stitutions and to have civic skills. There may be no
high school government, high vocabulary scores, or causal relationship between skill-acts and civic skills,
the ability to speak English have the capacity to and we should not draw causal arrows between
communicate in political settings or to organize po- skill-acts and civic skills as we do on Figure 3. Finally,
litical activities. A similar but more severe problem there is the locus of development problem, the issue of
applies to civic skills developed in adulthood. It where skills are developed. Even if the correlation
seems likely that respondents who report engaging between skill acts and civic skills is not the result of
in an activity such as writing a letter or organizing a some third variable, the relationship may be due to
meeting have the skill to do so. However, the con- the fact that civic skills lead to the performance of
verse is not necessarily true; all those with the skill to skill-acts, not vice versa. Individuals may perform
write letters do not necessarily do so at work. Our skill-acts in a particular institutional setting because
measure of adult civic skills, then, really measures they brought skills with them, perhaps innate skills
engaging in activities that require skills, what we or skills learned elsewhere. If this is the case, civic
shall call skill-acts. Our claim that skill acts measure skills will not be developed in nonpolitical institu-
civic skills amounts to arguing there must be at least tions through skill-acts. The exercise of skill-acts in a
some correlation between civic skills and the three particular institutional setting will not imply that civic
skill-acts variables at the center of Figure 3. skills are obtained or enhanced in that setting.
Second, correlation is not enough. Even if some In short, to establish the proposed links between
measure of civic skills is correlated with skill-acts, civic skills and political participation we must dem-

277

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 A Resource Model of Participation June 1995

onstrate that our measures indeed measure real skills; column. For example, such preexisting civic skills as
that there is no unspecified additional factor (e.g., a the ability to write a letter combine with opportuni-
taste for involvement) responsible for both the per- ties on the job to write letters to produce job skill-acts
formance of skill acts and political activity; and that captured by our questions. In turn, these skill-acts (as
these skills were developed or enhanced in the non- shown by the arrow going from skill-acts to civic
political institutions where they are exercised. skills) develop and enhance civic skills.
One response to the measurement problem is that As indicated by the rounded boxes on Figure 3, we
our measures have a good deal of face validity. This is do not have direct measures for several of the key
certainly the case for education and language ability. variables, skill opportunities, and civic skills. There-
It also applies to the exercise of skills in adult insti- fore, we must find proxies for them. Whether an
tutions. Because those who perform a skill-act are individual gets the opportunity to practice a civic skill
presumably learning a new skill or maintaining and at work, in an organization, or at church depends
improving a preexisting one by practicing it, the upon several things. Obviously, it depends upon the
measure of skill-acts is very likely to be an indicator of institutional involvements listed in the second column
the existence of civic skills.24 Moreover, because we in Figure 3-having a job or being affiliated with a
ask about skill-acts in three major secondary institu- voluntary association or religious institution. We
tions, we capture the main opportunities people have measure workplace involvement by a three-point
to practice such skills. Despite these arguments in scale of employment status (not working scored as 0,
their favor, the questions about skill-acts have what part-time as 1, and full-time as 2); attachment to
might seem at first to be a disconcerting feature. organizations by a three-point scale (ranging from no
Responses to them are not all highly correlated across involvement, scored as 1, to attending meetings,
the three domains of job, organization, and church.25scored as 3); and involvement in religious institutions
How can these questions measure a coherent one- by a nine-point measure of frequency of religious
dimensional concept like civic skills if a person who attendance and a six-point measure of the number of
reports performing skill-acts in one domain is not hours devoted to church activities. Beyond institu-
very likely to do so in another? Answering this tional involvements, the opportunity to practice skills
question requires some careful modeling of how also depends on the type of institution listed in the first
opportunities to exercise skills interact with the skills column in Figure 3. Occupations, voluntary organi-
individuals already possess to produce the kinds of zations, and churches differ substantially in the ex-
skill-acts our respondents report. tent to which they afford opportunities to exercise
With some extensions, a model can also provide skills germane to politics. Someone who works in a
the basis for meeting the other two problems we have consulting firm rather than a dry-cleaning shop, who
delineated. We must show that the performance of is involved in a fraternal organization rather than
skill-acts in each domain represents an exercise of civic softball league, or who is active in a congregationally
skills (not merely the reflection of some third vari- organized, rather than a hierarchically organized,
able) and that those who perform skill-acts thereby church is more likely to have an opportunity to
develop civic skills as well. In other words, people use develop civic skills. For jobs we describe the type of
preexisting civic skills (education-based organiza- institution by a nine-point occupation scale in
tional and communications skills as well as innate which higher values indicate higher-status jobs,
skills) to perform skill-acts. In turn, when they per- which presumably provide more opportunities for
form skill-acts in one institution they increase their skill-acts, and for religious institutions we use a
skills so that they can engage in still more skill-acts in dummy variable for belonging to a Catholic church,
that or some other domain. Establishing this kind of which provides fewer opportunities for skill-acts,
reciprocal causation is usually very difficult in cross- presumably because of its hierarchical structure.
sectional studies without making some arbitrary "ex- Obviously, it would have been preferable to have
clusion" restrictions on which variables affect other direct measures of civic skills, but we could hardly
ones but we have leverage on this problem because follow our respondents around to observe their skill
we have asked people about their activities in three in making public statements or organizing meetings.
different domains that cover most of the major op- As substitute measures, we use our respondents'
portunities adults have to gain skills. Within each subjective reports of their civic skills. Self-described
domain, it is obvious that the type of institution and skills (the box in the upper right-hand corner of Figure
the level of involvement in it should affect skill-acts in 3) are measured by questions as to whether respon-
that domain but not in the other domains. This dents believe they could write a convincing letter on
provides some obvious exclusion restrictions. a public matter, could talk well at a meeting, and
Figure 3 displays what we believe to be the recip- would be taken seriously if they made a public
rocal causation between civic skills and skill-acts in statement. These three items are correlated about .32
the three domains of work, organizations, and with each other and form a reasonable eight-point
church. We propose that an individual's preexisting scale. There are, however, potential problems with
civic skills represented by the shaded rounded box at this measure. For one thing, it raises the same ques-
the upper right of Figure 3 combine with opportunities tions we discussed earlier with respect to measures of
to practice skills in each domain listed in the third political efficacy (which, in fact, it resembles); that is,
column to produce the skill-acts listed in the fourth high self-assessments of civic skills could be the result

278

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 American Political Science
as well as the cause of political activity. This limits the
usefulness of the measure for explaining political
participation but does not affect its usefulness for
understanding the relationship among civic skills,
skill-acts, and opportunities to exercise skills.27 A
second potential defect is of greater concern: people's
self-assessments are unlikely to be fully accurate as
measures of their real talents. At worst, however, the
measurement error inherent in respondents' subjec-
tive reports of their civic skills would probably de-
press our estimates of the relationships between civic
skills and skill-acts. Therefore, in this context, self-
described skills are a reasonable measure of civic
skills.
Estimates of the learning model for civic skills are
in Appendix B. We show that there is a single
dimension of "civic skills" underlying the various
measures of civic skills-the three measures of adult
skill-acts, the two educational experiences, and lan-
guage ability. Moreover, people develop civic skills
through their involvement in the institutions of adult
life. Engaging in skill-acts (planning meetings, mak-
ing speeches, etc.) develops civic skills that are po-
tentially transferable to politics. We also show that
the nature of the institution itself affects the number
of skills exercised there-even after a number of
individual characteristics that might affect the perfor-
mance of skill-acts have been taken into account. This
supports the notion that we are observing a real
process of skill development within institutions, not
merely the consequences of the attributes that people
bring with them. The analysis in Appendix B dem-
onstrates in a concrete way how churches, jobs, and
organizations can serve as potential training grounds
for political activity and how nonpolitical choices
about jobs, organizations, and church attendance
may affect political life. To demonstrate that they do,
in fact, have such an effect is our next, and most
fundamental, task.
POLITICAL PARTICIPATION
AND RESOURCES
We have shown that accidents of birth combine with
choices about jobs, family, and organizational and
religious involvement to determine the resources of
time, money, and skills that individuals bring to
politics. We now inquire about the links between
these resources and political activity. We construe
political participation quite broadly and include the
following in the scale:
voting in the 1988 presidential election (70% of the
population)
contacting at federal or local level in last year (34%)
giving campaign money between January 1988 and
the interview in Spring 1990 (24%)
working informally with others on community prob-
lems in the last year (17%)
campaign work between January 1988 and the inter-
view in Spring 1990 (9%)
279
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
Review Vol. 89, No. 2
protesting during the last two years (5%)
board membership or regular meeting attender dur-
ing last two years (5%)
These participatory acts vary along many dimen-
sions. Some require the investment of time, some
money. Some require skill, others do not. Some are
ongoing, others episodic. And some are considered
mainstream, others less so. Our survey has questions
about several other activities such as voting in local
elections, participating in political organizations, and
phoning in to radio talk shows, but the activities
listed above cover the major dimensions of political
activity. Our survey also includes detailed informa-
tion on the time and money devoted to political acts.
Later we will use this additional information, but for
the moment, we score each of these seven activities
as 1 for people who engage in it and 0 otherwise.
Then a simple sum28 of the number of acts yields an
average of 1.63 activities in the sample (with a stan-
dard deviation of 1.35). This suggests that the aver-
age person votes and performs part of another act. In
fact, the frequency distribution of acts is skewed with
21% performing no acts, 33% one, 21% two, 14%
three, 7% four, 3% five, and 1% or less six or seven
acts.29 Thus, three-quarters of the population per-
forms between zero and two acts, and the remaining
quarter is concentrated at three or four acts.
Estimating the Model
We presented the complete resource model in Figure
3, in which such resources as free time, family in-
come, and civic skills, along with citizenship status
and political interest, explain the level of political
participation. We do not observe civic skills directly
but have shown that they can be represented by
skill-acts, language abilities, and formal educational
experiences. This suggests that we can use ordinary
least squares (OLS) to regress political acts on free
time, family income, skill-acts, language abilities, and
formal educational experiences. Proceeding in this
way, however, depends upon our faith in the exoge-
neity and reliability of the resource measures and our
willingness to make assumptions about what affects
political participation. We have argued that one of the
advantages of a resource explanation for political
participation is that resources result from decisions
about life circumstances, jobs, joining organizations,
and attending church that are temporally prior to
political participation. This seems obviously true for
family income and free time, and it seems very likely
true for adult civic skills developed at work, in
church, and in nonpolitical organizations. If so, re-
sources can be considered exogenous and OLS can be
used to estimate the resource model.
This leaves us with one possible problem. If some
omitted variable (e.g., a taste for participation) is
correlated with both the accumulation of resources
(most likely civic skills as measured by skill-acts) and
political participation, then ordinary least squares
will yield biased estimates. There are good reasons to
 A Resource Model of Participation June 1995

believe that this might not be much of a problem in

our model. First, the inclusion of participation in high
school governance probably helps control for pre- Determinants of Overall Political Participation:
existing tastes for political activity. Second, we do not Ordinary Least Squares Estimation
believe that the process that leads to involvement in a
OVERALL POLITICAL
church, for example, has much to do with a taste for
PARTICIPATION MEASURE
political participation. Yet we do believe that psycho-
logical engagement with politics matters for political MODEL WIO MODEL WITH
INTEREST INTEREST
participation, so that it seems sensible to add an
additional control for tastes in the participation equa- INDEPENDENT COEFF. BETA COEFF. BETA
tion by including the sum of interest in local and VARIABLE (SE) WT. (SE) WT.
national political affairs.30 This has the virtue of Political Interest - .261** .304
allowing us to compare the relative importance of .015
resources and psychological engagement. It has the Adult skill-acts
defect, however, that it may introduce another prob- Job .087** .101 .057** .066
lem. As we have discussed, political interest may be (.022) .021
as much an effect as a cause of political activity: it may Organizational .137** .106 .123** .095
be endogenous. This may require using two-stage (.029) .027
least squares (2SLS) to correct for biases created by Church .118** .088 .096** .072
the endogeneity of political interest. (.033) .031
Time and money
Ordinary Least Squares Estimates Free time .000 .002 .004 .013
(.007) .006
Before going to 2SLS estimates, it is instructive to Family income .051** .112 .047** .104
consider the OLS estimates in Table 2, which reports (.009) .008
regression coefficients, standard errors, and beta Institutional involvements
weights3' for the impact of various resources in a Working - .045 - .030 - .008 - .006
linear model. 3 The only difference between the left- (.038) .036
and right-hand set of columns is the inclusion of Retired .388** .090 .313** .073
political interest in the latter equation. The results (.090) .085
clearly demonstrate the importance of resources. Ex-
Organizational .070 .043 .031 .019
cept for speaking English at home, the measures of (.036) .034
civic skills acquired early in life (education, participa-
Attendance at .010 .021 .001 .002
tion in high school government,33 and vocabulary
church (.011) .010
ability) are positively related to political activity.
Time in church .049 .043 .053 .047
When citizenship is left out of the equation, speaking
work (.028) .027*
English at home appears to have an impact, but our
analyses suggest that this is merely because it spuri- Institutional types
Occupation .020 .040 .021* .042
ously picks up the impact of citizenship with which it
(.01 1) .010
is correlated at .48. Citizenship must be included in
the equation because it is a prerequisite for voting Catholic church .061 .020 .086 .028
(.055) .052
and might affect other kinds of participation as well.
Civic skills acquired as an adult in nonpolitical insti- Formal education
tutions are also significant, making clear the role of Years of .145** .164 .120** .136
education (.021) .020
the social institutions of civil society in creating a
competent and active citizenry. Family income also High school .178** .130 .1 18** .086
matters a great deal. The only resource measure that governance (.025) .024
does not have a significant effect is free time. As we Language ability
shall see, when we purge free time of error, it too Speaking English .045 .011 .056 .014
emerges as a significant factor in explaining political at home (.077) .073
activity. Vocabulary score .062** .099 .032* .051
Now consider the right-hand columns reporting (.013) .012
the equation that includes political interest. Note that Citizenship .889** .109 .790** .097
with the exception of retirement, neither the degree (.158) .150
of institutional involvement nor the type of institu- Constant -1.380** - -2.281**
tion is more than modestly important in determining (.193) .190
participation. Time spent in educational, charitable, R2 .301 .377
or social activities associated with a church has a Sample size 2,438 2,429
small impact that is barely statistically significant but
Source: Data from Citizen Participation Survey.
none of the rest of the institutional involvements has Note: CoEFF. refers to the regression coefficient and SE to its standard
a statistically significant impact on political participa- error. BETA WT. refers to the standardized regression coefficient.
?p .05.
tion. Simply being involved with nonpolitical institu-
**p ? .01.

280

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 American Political Science Review Vol. 89, No. 2

tions does not foster political activity. What counts is OLS results converge on a common interpretation in
what happens there-in particular, whether there are which resources and psychological engagements
opportunities to learn, improve, or maintain skills. drive political participation.
This result elaborates in an important way our under-
standing of the role of nonpolitical institutions in
Two-Stage Least Squares Estimates
stimulating political activity.
That political interest is related to political activity
Whereas the OLS model incorporates the possibility
is, in itself, not very illuminating. As we have pointed of direct effects from institutional involvements and
out, political interest is likely to be a consequence as types to political participation, our first 2SLS model
well as a cause of political activity. Even if we assumes that these paths are zero. We assume this
consider it only as a cause of activity, however, it is because our resource model suggests that these paths
hardly astonishing that those who are interested in should be zero. This provides a substantial set of
politics are also active.m What is important is thatvariables
its that were in the OLS equations but are
inclusion does not supplant the effects of civic skills excluded from the 2SLS model and used as instru-
and family income. The relationship between re- mental variables: working, retired, occupation, organiza-
sources and activity remains even with this powerful tional involvement, attendance at church, time devoted to
predictor of political activity in the equation. church work, and Catholic. In addition, we include in
the equation (and treat as exogenous) family income,
participation in high school governance, speaking English
Problems of Endogeneity
at home, formal educational experience, vocabulary score,
There is still room for worry, however. If political and citizenship status. This means that the three skill-
interest is endogenous, then the OLS estimates may acts measures and free time are treated as endoge-
be biased. Even if political interest is not endogenous, nous. Finally, we also use a set of individual charac-
OLS estimates may still be biased if some unmea- teristics that are clearly exogenous as instruments:
sured variable not captured by political interest af- race (African-American and all others), ethnicity (Lat-
fects both the accumulation of resources and political ino and all others), parents' education average on a
participation.35One approach to this problem (Achen nine-point scale, gender, number of children at home,
1986) is to find exogenous variables that explain whether any children are of preschool age, and
skill-acts but are not also proxies for the tastes that whether the spouse works full-time, part-time, or not a
directly cause political participation. These can then all.37 For the equation with political interest in it, we
be used as instrumental variables to purge the mea- use one additional instrument: the respondent's po-
sures of civic skills of this taste factor. litical interest as reported on a screener interview
Two-stage least squares is the most efficient completed 6 to 12 months before the final interview.38
method for combining the instrumental variables that Table 3 presents data from a 2SLS analysis in which
we need for civic skills and political interest. It is also institutional involvements and types (as well as the
useful for correcting for error in free time and in- set of demographic attributes and life circumstances
come.36 It requires the availability of good instru- listed) are used as instruments but excluded from the
ments-exogenous variables that are highly corre- equation. This approach should avoid the problem of
lated with the included endogenous variables. It also possible bias in the OLS estimates. The result for the
requires that enough instruments be excluded from resource model are very similar to what we found in
each equation to produce identification. Because we Table 3 except that free time, now that it is purged of
have a theory in which institutional involvements error, is also significant. Somewhat surprisingly, po-
and institutional types have no direct impact on litical interest and adult civic skills matter even more
citizen participation (see Appendix B), we can use in the 2SLS equation than in the OLS version. We
these variables as instruments in the participation believe this is because political interest and adult civic
equation. After all, our theory says that institutional skills are measured with error and 2SLS corrects not
involvements and types should be omitted from the only for endogeneity but also for unreliability.39 In
participation equation. One might, however, wonder this case, it appears that unreliability has depressed
whether institutional involvements might also proxy the OLS estimates.
a taste for participation. This suggests that it would The similarity of the coefficients for the three kinds
be interesting to include these measures in the equa- of adult civic skills is especially striking. If the skill-act
tion to see if they have an impact beyond resources. measures really measure skills and if the resources
Unfortunately, this may leave us with a meager and model is correct, then we would expect that skills
weak set of instruments. exercised at work, in nonpolitical organizations, and
Our solution is to proceed in two complementary in church would be transferable to politics at about
ways. We use 2SLS to estimate a model in which the same rate. If either of these hypotheses fail, it
institutional involvements are excluded and used as would seem unlikely that the coefficients would be
instruments. Then we use 2SLS to estimate a model equal to one another. A test for equality of the
that excludes institutional type and religious atten- coefficients strongly supports the conclusion that the
dance to ensure identification but that includes mea- three coefficients can be treated as equal to one
sures of institutional involvement for jobs, organiza- another.40 We impose this restriction on the skill-acts
tions, and church. These two approaches and the in the third column, and this shows that simply

281

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 A Resource Model of Participation June 1995

Although we do not report the details here, we

have also estimated this model with institutional
Determinants of Overall Political Participation: involvements (working, retired, organizational in-
Two-Stage Least Squares Estimation volvement, attendance at church, and time in church
activity) included as independent exogenous vari-
OVERALL POLITICAL
ables meant to represent tastes for participation. The
PARTICIPATION MEASURE
sum of skill-acts is still highly significant, with a
WITH ADULT WITH SUM t-statistic of 3.46 and a beta weight of .394. Moreover,
CIVIC SKILLS OF ADULT
the coefficients for the institutional involvements in
SEPARATE CIVIC SKILLS
the equation are, with the exception of the effects of
INDEPENDENT COEFF. BETA COEFF. BETA being retired and working, insignificant or of the
VARIABLES (SE) WT. (SE) WT. wrong sign.41 This suggests, once again, that it is
Political interest .420 .489 .399** .465 civic skills that matter for political participation, not
(.030) (.031) institutional involvements. In sum, our several
Aduft skill-acts modes of estimation show political resources to be
Job .163** .189 potent for political activity.
(.047) We can still conjure up other possible nuisances
Organizational .091** .070 that could explain our results, but we believe that we
(.040) have tried systematically to eliminate as many alter-
Church .177** .131 native explanations as possible with the data and
(.036) techniques at hand. We show, for example, that
Sum of adult - - .154** .303 skill-acts measure civic skills, that skills are devel-
civic skills (.021) oped in adult institutions, that civic skills predict
participation even with controls for political interest
Time and money
Free time .044** .150 .042** .142 and other confounding variables, and that a theoret-
ically generated set of instrumental variables produce
(.012) (.008)
the same result. In addition, we have tried systemat-
Family income .037** .082 .032** .072
(.009) (.008)
ically, with zeal and thoroughness, to make our
results evaporate by estimating many other models.
Formal educational
For example, we have included age, length of resi-
experiences
Years of .089** .101 .079** .090 dence, intensity of party identification, and many
education (.023) (.020) other variables in other specifications not reported
here, and tried many nonlinear specifications. The
High school .073** .053 .070** .051
governance (.026) (.025) results we report are typical of what we get; and in no
case have we been able to eliminate-or even reduce
Language ability
much-the strong and significant results of civic skills
Speaking English .034 .009 .039 .010
at home (.076) (.074) and money.

Vocabulary score .010 .016 .011 .017

(.013) (.013)
U.S. citizenship .699** .085 .686** .084
(.156) (.153)
Constant -2.862** -2.721 **
(.202) (.200)
R 2 .337 .345
Sample size 2,427 2,430
Source: Data from Citizen Participation Survey.
Note: CoEff. refers to the regression coefficient and SE to its standard
error. BETA wr. refers to the standardized regression coefficient. Instru-
mental variables for 2SLS estimation are working at job, retired or not,
occupational type, degree of organizational involvement, attendance at
church, time in church activities, Catholic, years of education, involve-
ment in high school governance, speaking English at home, vocabulary
score, family income, black, Hispanic, education of parents, number of
kids, preschool kids, sex, spouse work full-time, spouse work part-time,
citizen, and interest in politics from the screener. The endogenous
variables are therefore political interest, job skill-acts, organizational
skill-acts, church skill-acts (or the sum of these three), and free time.
Up s .05.
**p S .01.
taking the sum of skill-acts yields a highly significant
coefficient with a beta weight (.303) roughly compa-
rable to the impact of political interest (.465).
POLITICAL RESOURCES AND
SPECIFIC POLITICAL ACTS
The resource model works very well for an overall
measure of political participation. However, a sum-
mary activity measure presumably masks significant
differences among political acts, differences that
might be related to resources. We distinguish three
kinds of activity: acts requiring an investment of
money (through contributions to campaigns and po-
litical causes); acts requiring an investment of time
(by, e.g., working in a campaign or on a community
issue, taking part in a protest, contacting an official);
and voting. We would expect these to differ in their
resource requirements. Making a contribution obvi-
ously demands money but should require little in the
way of free time and may not require skills. The
time-based acts obviously demand some free time
and probably require, on average, a higher level of
civic skills. As the easiest political act, voting ought
not to require much in the way of resources except,
perhaps, some free time to get to the polls. Political

282

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 American Political Science
interest might be especially important for voting
because there are so few concrete payoffs to voting.
Voting
We begin with what is seemingly the least demand-
ing form of political activity, voting. To construct the
dependent variable, we combine two questions-one
each about how often the respondent votes in local
and national elections-to yield a nine-point scale. To
simplify the presentation of the effects of the inde-
pendent variables, we combine the three adult civic
skills measures into one measure by simply taking
their sum for each person." The left-hand columns of
Table 4 presents the results of a 2SLS analysis in
which we regress the nine-point voting scale on
average adult civic skills and other variables in the
resource model. We use the same instruments as
before. Political interest has by far the most substan-
tial impact, with free time and citizenship also signif-
icant. (The beta weights are .54, .23, and .15,
respectively.) Income has no impact and civic skills
have less impact (a beta weight of .21) than they do
with the full participation index (where they have a
beta weight of .30).
A consistent finding in the literature is the substan-
tial impact of education on voting (Wolfinger and
Rosenstone 1980; Teixeira 1992). Consequently, the
statistically insignificant and incorrectly signed im-
pact of education on turnout is especially surprising,
but we believe that it is correct. Because past work
has not treated political interest as a possibly unreli-
able and endogenous measure, we believe it has
substantially underestimated the impact of interest
and overestimated the direct impact of education.
Indeed, we have replicated our results using National
Election Studies data and validated and self-reported
votes.46 This does not mean, however, that education
is unimportant. When we regress political interest on
formal educational experiences and language abili-
ties, we find that the impact of education on voting is
funnelled entirely through political interest. Educa-
tional attainment (beta = .17), participation in high
school government (.19), and vocabulary (.17) all
have an effect on political interest. (The effect on
political interest of speaking English at home (.03) is
much smaller.) In summary, our work leads to a
revised picture of voting as an act that is driven very
strongly by political interest and that requires little in
the way of money. Indeed, political interest is much
more important than resources if our main project is
to explain voting turnout.
Money
Our interest, of course, extends beyond voting. Mak-
ing political contributions is an increasingly im-
portant mode of citizen participation (Sorauf 1988).
When it comes to explaining contributions, the re-
source model provides striking results. We asked our
respondents whether they had made contributions to
electoral campaigns or, in response to direct-mail
283
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
Review Vol. 89, No. 2
requests, to any political cause. If they had, we asked
how much they gave. By adding the amounts given
to campaigns and causes, we develop a measure of
the amount donated to politics. The middle columns
of Table 4 present the results of a 2SLS analysis in
which this variable is regressed on the variables in the
resource model. The results are unambiguous: the
major determinant of giving money is having money.
Years of education also matter, but neither free time
nor civic skills' affect monetary contributions. Strik-
ingly, even political interest has only a modest impact
(beta weight of .08) on the amount donated to poli-
tics. In short, it is easy to explain the amount given: a
contributor needs money-and little else in the way
of civic skills or political interest-to give money.
Time
The impact of resources on the forms of political
activity that require giving time (working in a cam-
paign, contacting government officials, protesting,
engaging in informal community activity, serving on
a local governing board or attending board meetings)
is quite different from the pattern we observed for
voting or monetary contributions. In the right-hand
column of Table 4, we present the results of an
estimation of the resource model with the dependent
variable as the number of time-based acts performed
by the respondent. Political interest clearly matters
(beta weight of .33), as does free time (beta of .09).
Family income does not matter, but civic skills clearly
have a significant impact. Adult civic skills exercised
in nonpolitical institutions18 (beta of .30) and partici-
pation in high school government (beta of .09) both
matter. Educational attainment has only a weak im-
pact, but this is because so much of its impact is
funneled through adult civic skills and political inter-
est. (Educational attainment is correlated at .48 with
the average of adult civic skills and at .33 with
political interest.)
The contrasting patterns for voting and performing
time-based acts, particularly in relation to skills and
interest, bear elaboration. Because formal education
simultaneously stimulates political interest and incul-
cates civic skills, both interest and skills have signif-
icant positive bivariate relationships with two forms
of participation: voting and performing time-based
acts. However, these equations demonstrate that
while voting appears to require interest but much less
in the way of civic skills, time-based acts depend on
civic skills as well as interest. In short, education
affects voting not so much by imparting skills as by
increasing political interest. In contrast, education
and participation in high school government have an
impact on the performance of time-based acts by
enhancing skills.
Free time is also worth more consideration. For
each of three of the time-based acts (working in a
campaign, getting involved informally on a commu-
nity issue or problem, and serving on a local commu-
nity board or attending its meetings), we asked
activists the number of hours they give to the activity
 A Resource Model of Participation June 1995

Determinants of Different Types of Acts (Two-Stage Least Squares Estimations)

VOTING POLITICAL MONEY ACTS TAKING TIME

(0-8) ($0-$5,500) (o05)C
INDEPENDENT COEFF. BETA COEFF. BETA COEFF. BETA
VARIABLES (SE) WT. (SE) WT. (SE) WT.

Political interest .884** .542 16.0* .081 .191** .326

(.065) (8.1) (.023)
Sum of adult civic skills .200** .209 -5.5 -.047 .103** .298
(.044) (5.5) (.016)

Time and money

Free time .129** .232 3.1 .045 .018** .091
(.018) (2.2) (.006)
Family income .013 .015 34.9** .341 .009 .031
(.018) (2.2) (.006)
Educational experiences
Years of education -.042 -.025 12.2* .060 .027 .045
(.044) (5.4) (.015)
High school governance .003 .001 5.4 .017 .081** .087
(.054) (6.6) (.019)
Language ability
Speaking English at home -.025 -.003 -13.3 -.015 .066 .025
(.174) (19.5) (.055)
Vocabulary score .058* .049 .8 .006 -.009 -.020
(.027) (3.3) (.009)
Citizenship 4.1 10** .147 26.1 .014 -.002 -.000
(.575) (40.3) (.115)
Constant -3.563** -236.5** - -1.168
(.593) (52.5) (.149)
R 2 .235 .138 .202
Sample size 2,322 2,430 2,430
Source: Data from Citizen Participation Survey.
Note: CoEPP. refers to the regression coefficient and SE to its standard error. BETA wr. refers to the standardized regression coefficient. Instrumental variable
for 2SLS estimation are working at job, retired or not, occupational type, degree of organizational involvement, attendance at church, time in church
activities, Catholic, years of education, involvement in high school governance, speaking English at home, vocabulary score, family income, black,
Hispanic, education of parents, number of kids, preschool kids, sex, spouse work full-time, spouse work part-time, citizen, and interest in politics fro
the screener. The endogenous variables are therefore political interest, sum of adult civic skills, and free time.
aNational and local.
bCampaign and mail.
CBoard or meetings, informal, campaign, contact, and protest.
As .05.
**p .01.

each week. When we limit the analysis to the 16% of the pattern for making political contributions is quite
our respondents who devote an hour or more per different. Income is determinative for the decision to
week to one of these activities (n = 393), we find a donate, as well as the size of the contribution.
very strong relationship between the total number of The different effects of political interest, civic skills,
hours given and the amount of free time available. time, and money on participatory acts provide part of
Roughly speaking, each additional hour of free time the explanation for the well-known multidimension-
per day leads to about one-third more hour of polit- ality of political participation (Verba and Nie 1972).
ical activity per week.49 Thus the amount of free time Because different acts require different kinds of re-
available seems especially important for the amount sources and more or less political interest, they form
of time people give to activities. What we observe, distinct clusters. Our model provides a way of ex-
then, is a two-stage process of political activation. plaining the existence of these clusters. It also dem-
Political interest and resources like adult civic skills onstrates why formal education is so highly corre-
have a major impact on the decision to participate (freelated with virtually every political act. Education
time has a minor impact as well), but constraints on affects political participation in at least two separate
free time control the amount of the time-based politi- ways: for some activities, especially voting, education
cal activity once this decision is made. Interestingly, instills political interest and participatory motiva-

284

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 American Political Science Review Vol. 89, No. 2
tions; for others, especially those that require time,
education leads to skills that facilitate activity.50
THE IMPACT OF RESOURCES
ON POLITICAL ACTIVITY
While it is beyond our scope here to describe the
many ways that differences in resources contribute to
different rates of participation across groups distin-
guished by income, race, ethnicity, or other charac-
teristics, we can give a few illustrations of the effects
on political activity of changes in resources.
Church Involvement. Consider an otherwise average
person who has no involvement with a religious
institution. If he or she joins a church and begins to
attend services weekly and to devote an additional
three hours a week to other church activity, this
person will perform approximately two more church
skill-acts (for the coefficients for church activity and
attendance at church, see Table A-1), which will
produce an increase of political acts from the average
of 1.63 to 2.0 (see the left-hand column of Table 3).
This increase represents over 25% of the standard
deviation (1.35) of the summary acts measure.
Income. Consider an otherwise average person
whose income goes up by $10,000. The middle col-
umns of Table 4 suggest that in consequence, this
person's political contributions will increase over 50%
from an average of $66 to $101.
Free Time. Finally, consider the impact of children. If
there are preschool children at home, a person loses
3Y2 hours of free time each day. This means, based
upon the preceding analysis, that among those peo-
ple who are already putting in at least an hour a week
in political activities, the addition of preschool chil-
dren will reduce their total time spent on informal
community activity, campaign work, board member-
ship, or attending meetings by about an hour per
week.5
In each of these cases, the ordinary changes that
people experience in their lives (joining a church and
attending a Bible-study class, getting a large raise, or
having a baby) affect the amount of political resources
available and thus lead to significant changes in
political participation. Most importantly from the
perspective of understanding patterns of participa-
tion, changes in life circumstances have different
impacts on time, money, and civic skills; and these
resources, in turn, have different links to each kind
political act.
CONCLUSION
The model developed here demonstrates that moti-
vations such as interest in politics are not enough to
explain political participation. The resources of time,Which of the categories on this card best describes the highest
money, and skills are also powerful predictors of
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
political participation in America. A model that in-
cludes resources has several advantages in explaining
political activity. Resources can be measured more
reliably than is possible with the motivations (e.g.,
efficacy or political interest) that often are used to
explain activity. Furthermore, they are causally prior
to political activity, deriving from home and school,
choices about jobs and family, and involvements in
nonpolitical organizations and churches. The civic
skills that facilitate participation are not only acquired
in childhood but cultivated throughout the life cycle
in the major secondary institutions of adult life. In
this way, the institutions of civil society operate, as
Tocqueville noted, as the school of democracy.
The resource model permits us to go beyond the
"standard SES model" in two ways. First, by moving
to a more general level and specifying the resources
derived from socioeconomic position that can be
applied to politics, the model establishes the mecha-
nisms that link SES to participation. In addition, by
moving beyond SES and encompassing other re-
sources not based on socioeconomic position (e.g.,
patterns of religious affiliation or involvement in
nonpolitical organizations), we move toward an un-
derstanding of the disparities in activity among po-
litically relevant groups distinguished by character-
istics (e.g., race, ethnicity, or gender) in addition to
SES.
Finally, the resource model illumines American
politics. We have seen that different resources are
differentially available to various politically relevant
groups and differentially critical for various kinds of
activity. To give a reductionist version of our find-
ings-political interest is especially important for
turnout; civic skills, for acts requiring an investment
of time; and money, for acts involving an investment
of money. To the extent that money is the least
equally distributed resource and to the extent that
making contributions has become in recent decades
an increasingly important citizen activity, the charac-
ter of American politics is profoundly altered.
APPENDIX A: THE QUESTIONS
SES and Social Characteristics
Income. Which of the income groups listed on this card includes
the total 1989 income before taxes of all members of your family living
in your home? Please include all salaries, wages, pensions, divi-
dends, interest, and all other income. (If uncertain,) What would
be your best guess?
Education.
of
What is the highest grade of regular school that you have com-
pleted and gotten credit for? (If necessary, add,) By regular
school we mean a school which can be counted toward an
elementary or high school diploma or a college or university
degree.
Did you get a high school diploma or pass a high school equiva-
lency test?
What is the highest degree you have earned?
educational level (mother/father) completed and got credit for?
285
 A Resource Model of Participation June 1995

Race and Ethnicity. Do you consider yourself Hispanic or Latino? D. given a presentation or speech
What is your race? E. contacted a government official?

Citizenship. Were you born in the United States? (If no,) Are you an Religious Organizations.
American citizen?
Now on a different subject, what is your religious preference? Is it
Work. Protestant, Catholic, Jewish, some other religion, or no religion?
What specific denomination is that, if any?
Last week, were you working full-time for pay, working part-time
Now I would like to ask you some questions about your religious
for pay, going to school, keeping house, or something else?
activity. How often do you attend religious services?
(Before you retired/In the last five years,) did you ever work for as
If you average across the last twelve months, about how many
long as one full year?
hours per week did you give to (church/synagogue) work-
What kind of work (do you/did you) normally do? That is, what
aside from attending services?
(is/was) your job called? (If necessary, probe,) What (are/were)
Here is a list of things that people sometimes have to do as part of
your main duties?
their church/synagogue) activities. After I read each one, please
Here is a list of things that people sometimes have to do as part of
tell me whether or not you have engaged in that activity in the
their jobs. After I read each one, please tell me whether or not
last six months as part of your (church/synagogue) activities.
you have engaged in that activity in the last six months as part
Have you
of your job. Have you

A. written a letter A. written a letter

B. gone to a meeting where you took part in making decisions B. gone to a meeting where you took part in making decisions
C. planned or chaired a meeting C. planned or chaired a meeting
D. given a presentation or speech D. given a presentation or speech
E. contacted a government official? E. contacted a government official?

Is your (husband/wife/partner) currently working part-time for

pay, going to school, keeping house, or something else?
Family Structure.
How many children do you have living at home with you? Please
include step- and adopted children living in your household.
Is this child under age 5? How many of these children are under
age 5?
Language. What language do you usually speak at home-English
or something else?
Vocabulary. Now we would like to know something about how
people go about guessing words they do not know. On this card
are listed some words-you may know some of them, and you
may not know quite a few of them. On each line the first word is
in capital letters-like BEAST. Then there are five other words. Tell
me the number of the word that comes closest to the meaning of the
word in capital letters. For example, if the word in capital letters is
BEAST, you would say "4" since "animal" comes closer to BEAST
than any of the other words. If you wish, I will read the words to
you. These words are difficult for almost everyone-just give me
your best guess if you are not sure of the answer.

Self-described Skills.
Institutional Involvements and Skills
Imagine you went to a community meeting and people were
High School Government. How active were you in school govern-
making comments and statements. Do you think you speak well
ment? Were you very active, somewhat active, not very active, or
enough to make an effective statement in public at such a
not at all active?
meeting?
If you did speak up, do you think people would pay a lot of
Time.
attention to what you said, some attention, very little attention,
About how many hours per day do you spend on necessary work or none at all?
for your home and family, including cooking, cleaning, taking Suppose you wanted to write a letter to someone in the govern-
care of children or other relatives, shopping, house and yard ment-perhaps your member of Congress or a local city offi-
chores, and so forth? About how many hours in total do you cial-on some issue or problem that concerned you. Do you feel
spend in an average day on such necessary activities for home that you write well enough to write a convincing letter express-
and family? ing your point or do you feel that you do not?
About how many hours do you spend on gainful employment in
an average day, including commuting and work that you take
home? Political Interest and Political Activities
About how many hours do you spend studying for a degree or
enrolled in courses for a degree in an average day? Interest in Politics.
About how many hours of sleep do you average a night?
Thinking about your local community, how interested are you in
Organizational Involvements. local community politics and local community affairs?
How interested are you in national politics and national affairs?
Here is a list of organizations. Please read through this list and [Screener item] How interested are you in politics?
when you have finished, I'll have some questions. Are you a
member of
Voting.
Have you attended a meeting of the organization in the past twelve
months?
Does this organization sometimes take stands on any public In talking to people about elections, we find that they are some-
issues-either locally or nationally? times not able to vote because they're not registered, they don't
Here is a list of things that people sometimes have to do as part of have time, or they have difficulty getting to the polls.
their involvement with organizations. After I read each one, Think about the presidential elections since you were old enough
to vote. Have you voted in all of them, in most of them, in some
please tell me whether or not you have engaged in that activity
in the last six months as part of your involvement with this of them, rarely voted in them, or have you never voted in a
presidential election?
organization. Have you
Thinking back to the national election in November 1988, when the
A. written a letter presidential candidates were Michael Dukakis, the Democrat,
B. gone to a meeting where you took part in making decisions and George Bush, the Republican, did you happen to vote in
C. planned or chaired a meeting that election?

286

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 American Political Science Review Vol. 89, No. 2
Campaign Work. Since January 1988, the start of the last national
election year, have you worked as a volunteer-that is, for no pay
at all or only for a token amount-for a candidate running for
national, state, or local office?
Campaign Money.
Since January 1988, did you contribute money-to an individual
candidate, a party group, a political action committee, or any
other organization that supported candidates?
In your best estimate, about how much money in total did you
contribute since January 1988?
Contacting.
In the past twelve months, have you initiated any contacts with a
federal elected official or someone on the staff of such an official:
I mean someone in the White House or a Congressional or
Senate Office?
What about a nonelected official in a federal government agency?
Have you initiated a contact with such a person in the last
twelve months?
What about an elected official on the state or local level-a governor or
mayor or a member of the state legislature or a city or town
council-or someone on the staff of such an elected official?
And what about a nonelected official in a state or local government
agency or board? Have you initiated a contact with such a person
in the last twelve months?
Protesting. In the past two years, since (current month 1988), have
you taken part in a protest, march, or demonstration on some
national or local issue (other than a strike against your employer)?
Board Membership. Now some questions about your role in your Skills = e + d, (Skill-Acts 1) + d2 (Skill-Acts 2)
community. In the past two years, since (current month 1988), have
you served in a voluntary capacity-that is, for no pay at all or for
only a token amount-on any official issues such as a town council,
a school board, a zoning board, a planning board, or the like?
Attend Meetings. Have you attended a meeting of such an official
local government board or council in the past twelve months?
Informal Community Work. Aside from membership on a board or
council or attendance at meetings, I'd like to ask also about
informal activity in your community or neighborhood. In the past
twelve months, have you gotten together informally with or worked
with others in your community or neighborhood to try to deal with
some community issue or problem? (If you have mentioned this
activity elsewhere, perhaps in connection with your church or
synagogue, or an organization or local campaign, don't repeat it
here.)
APPENDIX B: A LEARNING MODEL
FOR CIVIC SKILLS
Production of Skill-Acts
People engage in skill-acts when they are presented with oppor-
tunities on the job, in organizations, or in church to write a letter,
make a speech, organize a meeting, or participate in a meeting and
when they have enough preexisting skills to respond positively to
the opportunity. If they lack either preexisting skills or opportuni-
ties to exercise them, then they cannot carry out the activity. For
each domain j, skill-acts are the joint result of skill opportunities Skill Opportunities j = fj (Involvement j )
and preexisting skills:
Skill-Acts] = aj + bj (Skill-Opportunities j)
+ cj (Skills) + error, (A-1)
where aj is a constant and bj and cj indicate how skill-opportunities
and preexisting skills are converted into skill-acts. This equation
shows why all three measures of skill-acts can be useful indicators
of skills even though they may not be highly correlated with one
another. To the extent that individual opportunities to exercise skills
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
in jobs, organizations, and churches are not highly correlated with
one another, skill acts will not be highly correlated even though
they are partly determined by individual skills. And there is no
reason to expect a high correlation among institutionally based
opportunities to engage in skill acts. Though skills are transferable
and may be carried from one institutional setting to another, there
is little reason to expect that an individual involved in one type of
institution (e.g., having a job) would be more likely to be involved
in a different one (e.g., to attend church).
If skills are converted into skill-acts at the same rate across all
three institutional domains, then all cl, c2, and c3 will be equal. This
might be true, but it will only show up in our data if we have
exactly the correct functional form for equation A-1. This seems
unlikely.52 It is asking too much for the cj to equal one another, but
it is not asking too much to have the c; differ significantly from
zero. If a c1 is zero, then skills would not be a cause of skill-acts in
that domain even though, as we shall see shortly, skills might still
be a consequence of skill-acts. It seems unlikely to us, however,
that skill-acts would produce civic skills without also being a
product of civic skills. Therefore, the first requirement for our
model is that all cj differ from zero.
Developing Skills
A central assertion of the resource model is that skills result from
a "learning process." Not only do people engage in skill-acts
because they have civic skills, they also develop skills because they
perform skill-acts. When they write letters or organize meetings,
people become more adept at these activities: their civic skills
increase. Language ability (vocabulary score and speaking English at
home) and formal educational experiences (educational attainment
and participation in student government) also provide skills:
+ d3 (Skill-Acts 3) + d4 (Language Ability)
+ d5 (Formal Educational Experiences) + error, (A-2)
where e is a constant and d, through d5 indicate how skill-a
language ability, and educational experiences are converted into
civic skills. This equation will almost certainly be misspecified if we
omit major institutional settings that create skills because we
would expect a correlation between skill-acts in that domain and
skill-acts in another domain. Thus it is of great importance that
workplaces, organizations, and, churches encompass the major
institutional settings where civic skills can be learned in adulthood.
If all skill-acts were turned into skills at the same rate, we would
expect dl, d2, and d3 to be equal. This seems unlikely, however, for
several reasons: the functional form is uncertain, our measures of
skill-acts miss the frequency with which people practice skills in
each domain, and we did not ask in detail about the skill-acts
performed. For example, we might expect those who practice
skill-acts on the job to be more likely to learn from them because
they spend so much time at their jobs. However, if the skill-acts
performed on the job are merely routine, they would be less
productive of skills. In any case, because we believe that engaging
in skill-acts develops skills, we expect dl, d4, and d3 to be positive
but not necessarily equal.
Skill Opportunities from
Institutional Involvements
Opportunities for individuals to perform skill-acts in an institution
depend upon institutional involvements and institutional types:
+ gj (Type j) + error, (A3)
where the intercept is zero to determine the mean of the unob-
served skill opportunities measure.
Defining the Variables
To estimate the equations for the development of civic skills
requires measures of skill-acts, skill opportunities in adult institu-
287
 A Resource Model of Participation June 1995

Determinants of Each Adult Civic Skill: Two-Stage Least Squares Estimations

ADULT CIVIC SKILLS

ORGANIZATION
JOB SKILLS-ACTS SKILL-ACTS CHURCH SKILL-ACTS
INDEPENDENT COEFF. BETA COEFF. BETA COEFF. BETA
VARIABLES (SE) WT. (SE) WT. (SE) WT.
Self-described skills .595** .561 .286** .337 .135** .200
(.045) (.034) (.020)
Involvements in institutions
Working at job .742** .434 - - -
(.034)
Retired from job -.178*
(.095)
Organizational involvement - - .731** .458
(.032)
Attendance at church - .063** .167
(.007)
Time in church work - - - .479** .564
(.015)
Institutional types
Occupational type .110** .187
(.012)
Hierarchical church - - - -.235** -.104
(.034)
Constant -3.746** - -2.630 - -.987**
(.255) (.189) (.133)
R 2 .433 .301 .484
Sample size 2,448 2,445 2,448
Source: Data from Citizen Participation Survey.
Note: COEFF. refers to the regression coefficient and SE to its standard error. BETA wr. refers to the standardized regression coefficient. Instrumental variables
for 2SLS estimation are working at job, retired or not, occupational type, degree of organizational involvement, attendance at church, time in church
activities, Catholic, years of education, involvement in high school governance, speaking English at home, vocabulary score, family income, black,
Hispanic, education of parents, number of kids, preschool kids, sex, spouse work full-time, and spouse work part-tine. The endogenous variable in all
three equations is self-described skills.
sP .05.
P .01.

tions, language ability, formal educational experiences, as well as a
measure of civic skills themselves. The performance of skill-acts in
each of three domains is measured by a five-point scale (0-4) as
described earlier.53 We did not measure skill opportunities di-
rectly, but use as proxies institutional involvements and types
described, along with all the other variables in our model, in the
main text.
Estimating the Equations
Equations A-1 and A-2 are basic to a model of skill development in
which skill-acts are the result of having skills and getting the
opportunities to exercise them; and skills, in turn, are the result of
engaging in skill-acts, language ability, and formal educational
experiences. They form a system of equations in which self-
described civic skills and skill-acts appear on both sides of the
equations; these measures, then, are endogenous. A standard way
to estimate equations of this sort is 2SLS (Hanushek and Jackson
1977). This requires finding exogenous variables excluded from
each equation that can be used as instruments. These are easy to
find in this system of equations. For the three skill-acts equations
formed by substituting equation A-3 into A-1, we can use the
measures of institutional involvements and institutional types that
are not in the current equation and the measures of language
ability and formal educational experiences. For example, for the
equation for skill-acts on the job, we can use the measure of
organizational affiliations, attendance at church, time devoted to
church activities, the dummy for a Catholic church, and the
measures of language skills and formal educational experiences.
For the single skills equation A-2, we can use all the measures of
institutional involvements and types as instruments. In addition to
the instrumental variables that arise naturally from the system of
equations, we also use a set of individual characteristics that are
clearly exogenous. These are race (African American and all
others), ethnicity (Latino and all others), family income in thou-
sands of dollars, the average of parents' education on a nine-point
scale, gender, number of children at home, whether any children
are of preschool age, and whether or not the spouse is working full
or part time.
Table A-1 reports the results of 2SLS estimations of the three
skill-acts equations formed by substituting equation A-3 into A-1.
The first row shows the impact of self-described skills on each kind
of skill-act. All three regression coefficients are highly significant,
ranging from .13 to .59. A change of 1.5 in self-described skills
(about one standard deviation) leads to .9 more job skill-acts, .4
more organizational skill-acts, and .2 more church skill-acts. These
are significant effects for self-described civic skills measured on
five-point scales, and they amount to .56, .34, and .20 standard
deviation changes ("beta weights") in job skill-acts, organizational
skill-acts, and church skill-acts respectively.
Not surprisingly, the degree of involvement in an institution
(e.g., working full-time, rather than part-time) has a lot to do with
engaging in skill acts. In addition, it is striking how much the type
of institution-working in a higher-status occupation or attending

288

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 American Political Science Review Vol. 89, No. 2

TwO-Stage Least Squares Estimates of Determinants of Self-Described Skills

RESPONDENT'S SELF-DESCRIBED CIVIC SKILLS

FIRST EQUATION SECOND EQUATION THIRD EQUATION

INDEPENDENT COEFF. BETA COEFF. BETA COEFF. BETA

VARIABLES (SE) WT. (SE) WT. (SE) WT.
Adult skill-acts
Job .216** .229 .236** .251
(.030) (.029)
Organizational .148** .126
(.046)
Church .043 .029 .093* .063
(.041) (.038)
Sum .146** .262
(.018)
Formal educational experiences
Years of education .080** .083 .102** .105 .1 08** .111
(.025) (.024) (.024)
High school governance .172** .114 .189** .126 .164** .109
(.029) (.028) (.029)
Language ability
Speaking English at home .119 .027 .120 .028 .129 .030
(.078) (.078) (.078)
Vocabulary score .149** .217 .152** .222 .150** .218
(.014) (.014) (.014)
Constant 4.065** 4.008** 4.006**
(.171) (.171) (.170)
R 2 .234 .231 .218
Sample size 2,448 2,448 2,448
Source: Data from Citizen Participation Survey.
Note: COEFF. refers to the regression coefficient and SE to its standard error. BETA wr. refers to the standardized regression coefficient. Instrumental variables
for 2SLS estimation are working at job, retired or not, occupational type, degree of organizational involvement, attendance at church, time in church
activities, Catholic, years of education, involvement in high school governance, speaking English at home, vocabulary score, family income, black,
Hispanic, education of parents, number of kids, preschool kids, sex, spouse work full-time, and spouse work part-time. The endogenous variables are the
adult civic skills (job skill-acts, organizational skill-acts, church skill-acts, and sum of skill-acts).
Up < .05.
p ' .01.

a Catholic church-matters for job skill-acts and church skill-acts
respectively. In fact, moving from a Catholic to a non-Catholic
church has about the same impact on church skill-acts as a
one-standard-deviation increase in self-described skills, and mov-
ing from the lowest to the highest rung on the nine-point occupa-
tional ladder has a greater impact than a one-standard-deviation
increase in self-described skills.
The results in Table A-1 show that self-described civic skills and
skill opportunities lead to skill-acts. This means that we can use
skill-acts as a rough measure of civic skills. It does not, however,
prove that people actually learn such skills through their involve-
ment in adult institutions.
Using the same instruments as in Table A-1, Table A-2 estab-
lishes this important point by showing that the three measures of
skill-acts help explain self-described civic skills. Consider the
results for the first equation. The coefficients for job skill-acts and
organizational skill-acts are highly significant. The coefficient for
church skill-acts is disappointing, but an examination of the
correlation matrix for the coefficients reveals that there is a high
correlation between the estimate for the impact of organizational
skill-acts and church skill-acts. The correlation of -.38 between
these two coefficients means that the estimation procedure found it
hard to distinguish one from the other. This suggests that if we
dropped one of them, the other one will "take up the slack." In the
second equation, we drop the organizational skill-acts variable,
and the impact of church skill-acts becomes much larger and
statistically significant. When we omit both the organizational and
church skill-act measures (not reported), the regression sum of
squares becomes significantly smaller than when we include both
in the first equation. An F-test rejects the notion that we should
drop both variables. Finally, when we return to the first equation
and do a Mtest to assess whether the coefficients for organizational
skill-acts and church skill-acts are equal, we find that despite the
apparent difference, we cannot reject the hypothesis that they are
identical.'4
Taken together, these results imply that the joint effect of
organizational skill-acts and church skill-acts is not zero and that
the two kinds of skill-acts have identical impacts. There is good
reason, then, to believe that all three skill-act variables are impor-
tant determinants of civic skills. The third equation imposes the
constraint that all three skill-acts have the same impact by substi-
tuting their average for each of them individually. An F-test
implies that this is too strong an assumption, but the R-squared
and other statistics suggest that it is not a bad approximation to the
truth.
Table A-1 shows how having skills leads to skill-acts, and Table
A-2 shows how engaging in skill-acts leads to having more skills.
Table A-2 contains an additional important finding. With the
exception of the measure of speaking English at home, the other
quite varied measures of civic skills (formal education, vocabulary
score, and participation in high school government) all relate
significantly to self-described skills. This gives us confidence that

289

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
 A Resource Model of
we are measuring a coherent phenomenon that can be called civic
skills.
Notes
The authors would like to thank the National Science Foun-
dation, the Spencer Foundation, the Ford Foundation, and
the Hewlett Foundation for generous support. Norman Nie
helped us design the Citizen Participation Study, and he
contributed to the initial version of this paper. The Citizen
Participation Study has also benefitted from the unflagging
and extremely able assistance of Tami Buhr, Nancy Bums,
Chris Downing, Stephen Haggerty, William Hoynes, Jane
Junn, Martin Petri, Kenneth Stehlik-Barry, and Liesbeth ter
Schure.
1. People also avoid politics because "they aren't allowed
to participate." This was a major reason for nonparticipation
in the United States at one time and it remains important in
many countries.
2. Using an impressive time series of surveys, Rosenstone
and Hansen (1993) emphasize the importance of mobilization
by political leaders, mobilization around issues, and mobili-
zation by political opportunities. Their data are especially well
suited for showing how the changing political environment
encourages or discourages political participation. Our data are
especially appropriate for describing the linkages from SES to
resources to participation. A complete theory of political
participation must synthesize both perspectives.
3. Our approach has strong affinities with resource mobi-
lization theory in sociology, although we concentrate upon
the resources available to individuals instead of the resources
available to social movement organizations. Social movement
theorists argue that organizations mobilize people by the
skillful use of resources; we argue that resources are a
necessary condition for people becoming involved in politics.
The two theories converge in their emphasis upon the crucial
role of resources for political participation and in their descrip-
tion of the relevant resources. McCarthy and Zald (1977), for
example, consider time and money and "skills in lobbying,
accounting, and fund raising" as the most basic resources (pp.
1224, 1234).
4. The SES-participation relationship is well documented;
the finding appears "with monotonous regularity" (Nagel
1987, 59). For citations of relevant literature, see Bennett and
Bennett 1986, 183-186; Conway 1991, 21-27; and Milbrath and
Goel 1977, 92. The SES-activity connection has been elabo-
rated in multiple ways, including (1) differentiation among
political acts, usually distinguishing between voting and
"more difficult" political acts, with the SES-participation
relationship more potent for the latter (Verba and Nie 1972;
Verba, Nie, and Kim 1978); (2) analyses of the links between
SES and activity, usually focusing on such motivating atti-
tudes as political interest and efficacy (Almond and Verba
1963; Barnes and Kaase 1979; Kaase and Marsh 1979; Verba
and Nie 1972; Verba, Nie, and Kim 1978) or on the different
impacts of components of SES (Rosenstone and Hansen 1993;
Wolfinger and Rosenstone 1980); (3) analyses of factors that
diminish the impact of SES, often focusing on the mobilizing
effects of organizational affiliations or such attitudes as group
consciousness (Olsen 1982; Verba and Nie 1972).
5. After noting that "remarkably little effort has been
devoted to explaining why certain resources matter and
others do not," Wolfinger and Rosenstone unpack SES into its
constituent parts and demonstrate that it is education, rather
than income or occupation, that has consequences for voting;
but they can go no farther than to say, "We cannot measure
all resources directly; instead, we infer them from the indi-
vidual's demographic characteristics" (1980, 9). This makes it
hard for them to explain exactly how education has an impact
on turnout. We shall follow their lead by elaborating re-
sources and extending the analysis to forms of participation
other than voting.
6. Huckfeldt and Sprague take a similar position in their
careful review of the literature on contextual effects in politics.
290
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
Participation June 1995
They express skepticism about "simple versions of the self-
selection argument. To what extent do people choose their
location in the social structure? How many of us really choose
our workplace colleagues, our coreligionists, our neighbors?
To the extent that we are able to exercise control over our
surroundings, do we use political criteria in exercising such
choice? Or do we choose a job because it pays well? A church
because our parents raised us in it? And then we take the
politics that accompanies the choice" (1993, 294). There is still
the possibility that some resources may be produced through
political activity. Although it is hard to see how political
activity would generate rather than consume free time, it can
produce more income (e.g., if activists gain patronage jobs or
favorable treatment for their businesses). Similarly, those
who take part in politics may develop political skills "on the
job," that is, learn and improve political skills through their
political activity. Although some activists do generate political
resources through their voluntary participation, politics is
such a small part of the lives of even the most active citizens
that most of their income and skills must be amassed outside it.
7. Stratification theory describes the distribution of class,
status, and power, and their consequences for the operation
of society (e.g., Bendix and Lipset 1966; Wright 1985). We
shall show how money, time, and skills flow from class and
status and how these resources are converted into political
participation.
8. Milbrath and Goel note: "No matter how class is mea-
sured, studies consistently show that higher class persons are
more likely to participate in politics than lower class persons....
This proposition has been confirmed in numerous countries"
(1977, 92; emphasis original).
9. Contrary to the logic of the theory, many people vote
and join organizations even though they cannot possibly
hope to recoup their costs through their gains to self-interest.
If the range of self-interested benefits is, as it must be,
expanded to encompass such psychic benefits as the satisfac-
tion of doing one's civic duty, then the theory becomes much
less potent.
10. Another rational choice approach dating back at least to
Downs (1957) argues that lower information and transaction
costs for well-educated people means that they will be more
likely to engage in politics. Not much has been done to
elucidate this approach. Our concept of civic skills, however,
explains why transaction and information costs might be
lower among the well-educated.
11. The Chicago school takes into account constraints on
time as well as money (Becker 1965, 1976), the production of
commodities by the household (Michael 1973; Michael and
Becker 1973), and investment in human capital (Becker 1975).
In this theory, individual behavior is constrained by income,
time, and household production capabilities. Household pro-
duction functions vary from person to person depending
upon accumulated skills, but production capabilities can be
improved through the investment of time and goods in
human capital. This is exactly the argument we make here.
12. The question was on total income from all sources for
the family. With the exception of the lowest of the 16
categories (which was scored at two-thirds of the range) and
the highest (which was scored at one-and-a-half), responses
were scored at the midpoint of the range in thousands of
dollars.
13. We would have been able to generate more precise data
if we had asked respondents to keep time-budgets instead of
asking about a typical day, but this would have been too
complicated and costly in a survey designed to cover a wide
range of concerns. In fact, the results based on our approxi-
mations accord very well with the results contained in the
literature on time use (e.g., Hill 1985). We did not ask about
"free time" directly because pretesting indicated that this
concept had no clear-cut meaning to respondents, whereas
the time spent working, doing household work, studying and
sleeping seemed meaningful to them. We concluded that
there would be much more error in an ambiguous query about
free time than in the total error across all of our easily
understood questions.
American Political Science
14. On the similarities and contrasts between time and
money, see Sharp 1981 and Mincer 1962.
15. There is precedent for considering the role of civic skills
in facilitating participation. Strate and his colleagues (1989)
demonstrate the importance of "civic competence" for voting
turnout. However, the variables included in their measure of
civic competence (e.g., attentiveness to politics and level of
political information) are explicitly political. Therefore, we
cannot be sure that they are not also a result as well as a
source of activity. Wald mentions the extent to which "con-
gregational organizations may serve as leadership training
institutes for people who lack other means of exposure to
organizational skills" and cites studies that find a strong
relationship between attendance at church services and elec-
toral turnout but not between religious attendance and other
forms of political activity (1992, 35). In his study of parish-
connected, non-Latino Catholics, Leege (1988) finds a rela-
tionship between parish activity and political activity and
discusses the potential of parish activity for developing the
kinds of skills we measure here.
16. Schooling affects participation in several ways: it fosters
values conducive to participation, broadens social networks,
and creates income-producing occupational opportunities.
We shall return to these themes. Here, we focus on the
skill-creating aspect of education.
17. The eight categories and their scores were (1) grammar
school or less, (2) some high school, (3) high school graduate
or GED, (4) some college, (5) college graduate, (6) some
graduate work, (7) master's degree, and (8) PhD. or profes-
sional degree. Not much is gained by cluttering our equations
with dummy variables for these categories, so we have used
this eight-point scale throughout.
18. Participation in high school governance might also
measure a "taste" for participation. A close relationship
between "tastes" and "skills" is predicted by George Stigler
and Gary Becker (1977), who argue that as people develop
skills in an area (e.g., baseball, music, politics), they will be
more likely to prefer the activity because they can derive more
and more pleasure from it at the same cost. Whether it
measures tastes or skills, participation in high school gover-
nance belongs in our model.
19. Thorndike and Gallup describe this test as a "test of
verbal intelligence.... [that assesses] the nature of past
earnings and not the ability to make novel adaptations"
(1944, 78-79). The mean of 6.20 (with standard deviation of
2.15) on our vocabulary score is close to the mean of 6.51 (with
standard deviation of 2.25) reported by Alwin for the 1989
GSS, which covered a slightly different population (1991, 628,
table 1).
20. The vocabulary test may not be a very good measure of
verbal ability for a very small fraction of our sample. The 1.8%o
of the sample who sometimes or always spoke another
language at home besides English or Spanish (and therefore
did not have the choice of being interviewed in their own
language) might have done better if they had been inter-
viewed in their native language. Our results, however, are
not affected by excluding these people so we have left them in
our analysis.
21. These facts suggest that the vocabulary score measures
something more than just schooling, but for our purposes the
exact relationship between vocabulary score and education is
not important. What is important is that the vocabulary score
allows us to control for verbal ability wherever or however it
has been obtained.
22. Voluntary associations vary substantially in the extent
to which they are involved in politics. We isolated affiliations
with nonpolitical organizations as follows. Respondents were
presented a comprehensive list of 20 kinds of voluntary
organizations (e.g., unions, professional associations, frater-
nal groups, block clubs, political issue organizations). For
each category for which the respondent indicated an organi-
zational affiliation, we asked a series of follow-up questions
about that organization (or, if more than one, about the one
with which the respondent was most involved). Among these
questions was whether the organization takes stands on
public issues. We consider any organization that does not take
291
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
Review Vol. 89, No. 2
stands on public issues to be nonpolitical. Unless otherwise
specified, we are referring in our discussion to these nonpo-
litical organizations.
23. There is an interaction with gender: in families where
both spouses work full-time, preschoolers reduce the hours at
home for a woman by 1.5 hours more than they reduce the
hours at home for a man.
24. The skill-act items have several attractive features con-
sistent with the conclusion that they measure the acquisition
of transferable skills. For one thing, in each domain the four
items form similar Guttman scales (with differences explain-
able by variations in chances to perform the skill from one
domain to another) indicating that they measure something
common across institutions, not something specific to a
particular institution. In addition, one activity-contacting a
government official-included in the list asked in each do-
main (though we did consider it not an opportunity to gain a
civic skill but an actual political act) does not scale with the
other activities, whether one uses Cronbach's alpha as a
criterion, loadings or commonalities in a factor analysis, or the
coefficient of determination for a Guttman scale. For example,
the lowest loading for the four items in factor analyses of each
domain is .527 while the loadings for contacting are .279, .446,
and .385. This too is consistent with the hypothesis that these
activities are a common set.
25. The correlations are .30 between job skill-acts and
organization skill-acts, .09 between job skill-acts and church
skill-acts, and .29 between organization skill-acts and church
skill-acts.
26. By using these scales in our regressions, we are assum-
ing a linear relationship between our dependent variables and
each one of them. This turns out to be close to the truth in
every case. We have tried numerous specifications where we
use dummy variables for each occupation, workplace status,
level of attachment to organizations, and every other inde-
pendent variable defined by a scale. The results are very close
to those reported here, and using the scales simplifies the
reporting of results.
27. All that matters for our purposes is that self-described
skills measure civic skills. Then our demonstration that self-
described skills are related to the exercise of skill-acts implies
that civic skills are related to skill-acts.
28. A count of activities is a simple and straightforward
measure that simplifies the initial presentation and analysis.
The extensive literature on the multiple dimensions of partic-
ipation (Verba and Nie 1972) suggests the importance of
disaggregating this simple measure, and one of the strengths
of the resource model is its ability to predict separate dimen-
sions of participation, such as who is likely to give money,
work in campaigns, or engage in some other activities. The
resource model does this by considering what is common
across people (i.e., resource constraints) while lumping to-
gether acts with varying issue content. As with standard
turnout models that typically do not take into account the
multifarious issues that impel voters to go to the polls (e.g.,
Wolfinger and Rosenstone 1980), we assume for the resource
model that it makes sense to consign to the error term the
many issues that might motivate people to participate. On
average, we suppose that these issue considerations are
uncorrelated with the resources that constrain or enable their
activity. Because we asked our respondents whether any
specific issue motivated their participation, we have, how-
ever, been able to show (in work available upon request) that
the resource model works for specific issue areas, as well as a
heterogeneous collection of them.
29. A number of papers have noted that respondents
typically overreport forms of participation like these because
of a social desirability bias "in which cognitive dissonance can
lead to a rather consistent distortion of memory in order to
reinforce continued perception of oneself as a good citizen"
(Cahalan 1968; see Anderson and Silver 1986; Hill and Hurely
1984; Katosh and Traugott 1981; Silver, Anderson, and
Abramson 1986; Volgy and Schwarz 1984; Weiss 1986). One
explanation for our results could be that these same biases
inflate reports of involvement in high school governance,
church attendance, and other activities leading to a spurious
A Resource Model of
correlation with political participation. We worried about this
possibility but rejected it for several reasons. First, social
desirability bias has to enter in a very specific way so that
some subgroups are more prone to it than others for it to
produce spurious results (Brady 1986). Second, the evidence
we have reviewed is consistent with the notion that social
desirability bias is a general human trait that is uncorrelated
with specific characteristics of the respondents. This means
that it would not bias our results at all (except for the intercept
of our regressions because of overall overreporting). Third, if
there are different forms of social desirability (some people
think that church attendance is desirable, others do not, then
social desirability may lead to an underestimation of the
importance of some factors because of what amounts to a
classic errors in variables situation. Fourth, our questions
were designed to maximize true recall and to minimize false
reporting by asking details of each act so that we have tried to
minimize social desirability bias to begin with. Fifth, the most
crucial part of our model is the relationship between skills and
participation; but if all of our skills are socially desirable, then
they should correlate because of this common fact. In fact, the
correlations are .30 between job skill-acts and organization
skill-acts, .09 between job skill-acts and church skill-acts, and
.29 between organizational skill-acts and church skill-acts.
The lowest of these numbers is the upper bound on the
possible amount of common social desirability bias. It is hard
to believe that a correlation of .09 could account for all of our
results. A detailed memorandum available from the authors
elaborates upon this points.
30. The sum goes from 2 (not at all interested in either local
or national affairs) to 8 (very interested in both). The Pearson
correlation of .54 between the two items compares favorably
with the correlation of .55 between interest in national politics
and a general interest in politics question on the screener.
(Interest in local politics and the screener question correlate at
.50.)
31. Beta weights are not a perfect way to measure the
importance of a variable (Achen 1982), but they are conve-
nient for making comparisons across variables.
32. An alternative to the linear form would be the assump-
tion that participation requires the interaction (or product) of
political interest and resources, but the logarithm of this
functional form would be a linear form in the logarithms of
each variable. It might be worth testing this functional form if
the measures in the equation had a known metric (e.g.,
quantities or prices in a typical economics problem); but
almost all the measures in the equation have an unknown
metric, so we cannot be sure that we should take their
logarithms. Our approach to this problem is to stick with the
simple linear form in the text. Another approach is to include
an interaction term of, say, political interest and the sum of
civic skills to see if it matters. In fact, it does, with a highly
significant t-statistic of 5.93; but we do not believe that this
really takes us beyond what we already know from the linear
formulation that can be considered the first term in a Taylor
series approximation of a product. Still another approach is to
try out many alternative functional forms including the loga-
rithm of participation, a Poisson regression, logarithms of
some of the independent variables (e.g., family income), and
dummy variable versions of scales for occupation, education,
work status, and many other variables. Our results remain the
same under these alternative specifications.
33. Measures of involvement in high school sports or
having taken high school civics courses, by the way, had no
impact on political participation or on the development of
civic skills as reported in Appendix B.
34. Nevertheless, when we consider various political acts
separately, we shall see that political interest is not uniform in
its impact on all modes of participation.
35. This describes a "triangular" system of causation in
which participation is explained by resources and some other
exogenous variables and resources, in turn, are explained
only by the other exogenous variables. It is well known that
triangular systems can be estimated consistently by OLS only
if the error terms are uncorrelated. If there is some unob-
served factor such as a taste for organizational activity that
292
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
Participation June 1995
affects both resources and participation, then the error terms
will obviously be correlated and the resources variables in the
participation equation will be correlated with the error term in
the equation, which leads to specification bias (Achen 1986,
chap. 2).
36. So far, we have argued that we must include political
interest in the participation equation and treat skill-acts and
political interest as endogenous. Family income and free time
also present special problems. It seems likely that our mea-
sures of these resources contain error, and this means that
OLS estimates of their coefficients will be biased in unpredict-
able ways. The conventional wisdom is that their coefficients
will be biased toward zero and that other "proxies" will pick
up some of their effect. This is probably often true, but other
things can happen (see Achen 1985). We solve this problem
for the family income measure by using the income question
on the screener to calculate the reliability of the income
measure. This is then used to construct a new measure of
income that is corrected for error. The family income measure
that we use in all the regressions here is the predicted value
from a regression of the follow-up measure on the screener
measure. This amounts to a correction for attenuation. (We
used the screener value for family income in the small number
of cases where only the screener variables was available.)
Although a reasonable measure, family income measured in
this way is clearly only an approximation of the money
available to an individual to use for political or other contri-
butions. To solve this problem for free time, we treat it as
endogenous and we use family size, preschool children, work
status, and other variables as instruments to purge it of error.
37. The computer code and the data for all of our runs are
available from the authors upon request and will be archived
at the ICPSR at the University of Michigan.
38. We model the simultaneity between political interest
and political participation by assuming that current participa-
tion depends directly upon current interest (and other factors,
of course), which, in turn, depends upon current participa-
tion and past interest. Past interest is measured by a question
on a screener interview 6 to 12 months before the final
interview. In this setup, even if both interest items are
measured with error, 2SLS provides consistent estimates for
the participation equation. We can supply this proof upon
request.
39. Another possibility is that interest and skills are picking
up the effects of variables that appear in the OLS estimation
but are omitted from the 2SLS version based upon our
theoretical assumptions. If our theory is right, then they
should be omitted. If our theory is wrong, then they should
be included, but this leaves us with no instruments for solving
the endogeneity problem. If endogeneity is not a problem, of
course, then the OLS estimation suggests that these variables
should be omitted. We are stymied, then, only in the case
when endogeneity is a problem and our theoretical assump-
tions are wrong.
40. The appropriate jZ test is described by Judge and his
colleagues, and the value of 1.92 with two degrees of freedom
(highly insignificant with a probability value of about .35)
strongly supports the null hypothesis that the coefficients are
equal to one another (1985, 614).
41. The impact of free time becomes insignificant when we
include dummy variables for working and retirement in the
equation because these two variables are so highly correlated
with free time. This suggests the fragility of this result, but it
also raises questions about how we should think about the
impact of work and retirement on participation. It seems
likely that work increases participation through the develop-
ment of skills (job skills and working are correlated at .605)
and decreases participation by reducing free time (free time
and working are correlated at -.626). These seem to be the
main routes by which working can have an impact so that
once job skills and free time are included in an equation for
participation, working (and retirement) should be excluded.
When we do include working in an equation with job skills,
free time, and many other variables, we find that it has a
negative coefficient and that free time appears to have no
impact. Yet we can think of no consequence of working, other
American Political Science Review Vol. 89, No. 2
than its reduction of free time, which would cause its coeffi-
cient to be negative. Moreover, the estimated coefficients for
working and for free time are correlated at .609 with one
another, suggesting that working is acting in place of free
time. Of all our results, the impact of free time is the most
fragile; but there is strong evidence (discussed later that free
time, at the very least, affects the number of hours given to
time-intensive participation once the decision to participate is
made.
42. Another test of our model would be to include mea-
sures of recruitment to politics and measures of psychological
engagement other than interest. In other work (Verba, Schloz-
man, and Brady 1995), we have estimated both OLS and 2SLS
versions of our models after adding political efficacy, partisan
strength, political interest, political information, and a mea-
sure of recruitment to politics. We found no substantial
change in our results.
43. In the language of econometrics (Granger 1990; Leamer
1990), we have investigated the "fragility" of our results
regarding civic skills and income and found that alternative
specifications lead to the same result. As noted in n. 41, our
results regarding free time are more fragile than the others.
44. A e test where the null hypothesis is the equality of the
three coefficients for the skills variables in the voting equation
yields a value of 3.36 which is not even significant at the .10
level so that we cannot reject the null hypothesis (judge et al.
1985, 614).
45. The impact of citizenship is, of course, to be expected.
We include it, rather than running the equation for citizens
only (which might seem appropriate), because it enables us to
see the impact of the resources over and above this obviously
potent variable. In addition, we want to compare the role of
citizenship in connection with voting with its impact on other
acts for which it is not a requisite.
46. This becomes apparent using the 1990-93 National
Election Studies panel. An OLS regression of self-reported
vote for 1992 on education and interest in the campaign (there
was no general interest question in 1992) yields beta coeffi-
cients of .22 and .34 respectively. Because interest in the
campaign is likely to be unreliable and endogenous, the 1990
campaign interest measure is used as an instrumental variable
for it, and then the beta weights go to .08 for education and
.90 for interest. (The regression coefficient on education is
barely significant at the .05 level in a sample of 1,242.) The
1990 campaign interest measure is an excellent instrument for
1992 campaign interest if the only problem with interest is due
to its unreliability. If campaign interest is also endogenous,
then we need a still better set of instruments like those used
here. Nevertheless, using lagged campaign interest at least
solves the reliability problem, and it must be considered better
than using simple OLS, which has typically been done in the
literature. We get similar results when we use self-reports
about voting in 1988 from the same data set with betas of .11
for education and .44 for campaign interest. Finally, because
self-reports of voting might be strongly driven by interest
whereas validated votes might not be (perhaps because of
some social desirability bias), we also used validated vote in
1988. (Our data file did not have validated vote for 1992.) The
results were a beta of .03 for education (with a standard error
larger than the regression coefficient) and .43 for campaign
interest.
47. A x2 test where the null hypothesis is the equality of the
coefficients for the three skills variables in the political money
equation yields a value of 1.53, which has a probability value
of about .50 (judge et al., 1985, 614). Moreover, a x2 test for all
three equaling zero yields 1.59 with three degrees of freedom
which has a probability value of about .60. Finally, the
t-statistic for the sum of the skill acts reported in Table 5 is
-1.00 with a probability value of .32. The evidence is over-
whelming that civic skills have nothing to do with giving
money.
48. A x2 test where the null hypothesis is the equality of the
coefficients for the three skills variables in the equation for
acts requiring time yields a value of 4.35, which has a
probability value of about .15, so that we cannot reject the null
hypothesis (Judge et al. 1985, 614).
293
This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms
49. A regression of the total number of hours spent work-
ing on a campaign, getting involved informally on a commu-
nity issue, and serving on a local community board or
attending its meetings for just those with nonzero time on the
independent variables of resources (time, income, the sum of
adult civic skills, participation in high school governance,
education, vocabulary score, speaking English at home),
political interest, and citizenship shows that only free time
matters. Skeptics might suggest that this result might be
produced by some artifact in the way we measure time.
Perhaps some people consistently under- or overestimate
how much time they spend on activities in politics and in
other parts of their lives. In fact, if this were the case, it would
create a negative correlation between free time (which is
measured as 24 minus the hours devoted to paid work,
household maintenance tasks and child care, school, and
sleep) and political activity.
50. The impact of education is broader still, affecting the
networks people are in, as well as the likelihood that they will
be in high-paying jobs. We explore these connections in later
publications.
51. In this case, it matters whether our average person is
male or female. Since a working woman loses more free time
than a man when there is a preschooler at home, the impact
on the amount of time she gives to time-based political
activities will be that much greater.
52. It seems unlikely for two reasons. First, it seems im-
probable that we have hit upon the ideal way to measure
skill-acts in our first try. There are probably some difficulties
comparing our measures across domains. Second, equation
A-1 is basically a "production function" for skill-acts from
skills and opportunities. This suggests alternative functional
forms. For example, skill-acts might be a function of the
product (or interaction) of skills and opportunities as in a
Cobb-Douglas production function, but by simply taking
logarithms we could get the linear form in the text. Knowing
as little as we do about the proper functional forms, it is better
to stay with the simple linear formulations.
53. Those not involved in an institution were scored at
zero. However, those scored at zero are not necessarily
completely without skills. Hence, our observation of skills is
censored by whether or not the person was involved in a
particular institution. One of the reasons for constructing the
model in the text is to overcome this difficulty by including the
major institutions in which someone might develop skills. In
this section only, we measure skill-acts in all kinds of organi-
zations; elsewhere we refer to nonpolitical organizations only.
For the job skill-acts we imputed some missing data. Descrip-
tion of the method and code used to impute these data can be
obtained from the authors.
54. This test involves asking whether the difference be-
tween the coefficients for organization skill-acts and church
skill-acts is zero. The test requires knowing the covariance
between the estimates of these two coefficients, ascertained
from their correlation and the standard errors for each coef-
ficient as reported in the table. The value for the Mtest is 1.45,
well short of statistical significance.
References
Achen, Christopher H. 1982. Interpretation and Analysis of
Regressions. Beverly Hills: Sage.
Achen, Christopher H. 1986. The Statistical Analysis of Quasi-
Experiments. Berkeley: University of California Press.
Almond, Gabriel A., and Sidney Verba. 1963. The Civic
Culture. Princeton: Princeton University Press.
Alwin, Duane F. 1991. "Family of Origin and Cohort Differ-
ences in Verbal Ability." American Sociological Review 56:625-
38.
Anderson, Barbara A., and Brian D. Silver, 1986. "Measure-
ment and Mismeasurement of the Validity of the Self-
reported Vote." American Journal of Political Science 30:771-
85.
A Resource Model of Participation June 1995

Barnes, Samuel H., Max Kaase. 1979. Political Action: Mass
Participation in Five Western Democracies. Beverly Hills: Sage.
Becker, Gary S. 1965. "A Theory of the Allocation of Time."
Economics Journal 75:493-517.
Becker, Gary S. 1975. Human Capital. 2d ed. Chicago: Univer-
sity of Chicago Press. Milbrath, Lester W., and M. L. Goel. 1977. Political Participa-
Becker, Gary S. 1976. The Economic Approach to Human Behav-
ior. Chicago: University of Chicago Press. Mincer, Jacob. 1962. "Labor Force Participation of Married
Bendix, Reinhard, and Seymour Martin Lipset, eds. 1966.
Class, Status, and Power: Social Stratification in Comparative Economics, by National Bureau of Economic Research.
Perspective. 2d ed. New York: Free Press.
Bennett, Stephen Earl, and Linda L. M. Bennett. 1986. "Po- Nagel, Jack H. 1987. Participation. Englewood Cliffs, NJ:
litical Participation." In Annual Review of Political Science,
vol. I, ed. Samuel Long. Norwood, NJ: Ablex Publishing.
Brady, Henry E. 1986. "The Perils of Survey Research: Inter-
personally Incomparable Responses." Political Methodology
11(3-4):269-291.
Cahalan, Don. 1968. "Correlates of Respondent Accuracy in
the Denver Validity Survey." Public Opinion Quarterly 32:
607-21.
Conway, M. Margaret. 1991. Political Participation in the United
States. 2d ed. Washington: CQ Press.
Downs, Anthony. 1957. An Economic Theory of Democracy.
New York: Harper & Row.
Duncan, Otis Dudley. 1984. Notes on Social Measurement:
Historical and Critical. New York: Russell Sage.
Granger, C. W. J., ed. 1990. Modelling Economic Series. Oxford:
Clarendon.
Hanushek, Eric A., and John E. Jackson. 1977. Statistical
Methods for Social Scientists. Orlando, FL: Academic.
Hill, Kim Quaile, and Patricia A. Hurley, 1984. "Nonvoters in
Voters' Clothing: The Impact of Voting Behavior Misreport-
ing on Voting Behavior Research." Social Science Quarterly
65:199-206.
Hill, Martha S. 1985. "Patterns of Time Use." In Time, Goods
and Well-Being, ed. F. Thomas Juster and Frank P. Stafford.
Ann Arbor: Institute for Social Research.
Huckfeldt, Robert, and John Sprague. 1993. "Citizens, Con-
texts, and Politics." In Political Science: The State of the
Discipline II, ed. Ada W. Finifter. Washington: American
Political Science Association.
Judge, George G., W. E. Griffiths, R. Carter Hill, Helmut
Lutkepohl, Tsoung-chao Lee. 1985. The Theory and Practice of
Econometrics. 2d ed. New York: Wiley.
Kaase, Max, and Alan Marsh. 1979. "Political Action Reper-
tory: Changes over Time and a New Topology." In Political
Action, ed. Samuel Barnes, Max Kaase et al. Beverly Hills:
Sage.
Katosh, John P., and Michael W. Traugott. 1981. "The Con-
sequences of Validated and Self-reported Voting Mea-
sures." Public Opinion Quarterly 45:519-35.
Leamer, Edward E. 1990. "Sensitivity Analyses Would Help."
In Modelling Economic Series, ed. C. W. J. Granger. Oxford:
Clarendon.
Leege, David C. 1988. "Catholics and Civic Order: Parish
Participation, Politics, and Civic Participation." Review of
Politics 50:704-36.
McCarthy, John D., and Mayer N. Zald. 1977. "Resource
Mobilization and Social Movements: A Partial Theory."
American Sociological Review 82:121241.
Michael, Robert T. 1973. "Education in Nonmarket Produc-
tion." Journal of Political Economy 81(pt. 1) :305-27.
Michael, Robert T., and Gary S. Becker. 1973. "On the New
Theory of Consumer Behavior." Swedish Journal of Economics
75:378-96.
tion. 2d ed. Chicago: Rand McNally.
Women: A Study of Labor Supply." In Aspects of Labor
Princeton: Princeton University Press.
Prentice-Hall.
Olsen, Marvin E. 1982. Participatory Pluralism. Chicago: Nel-
son-Hall.
Rosenstone, Steven J., and John Mark Hansen. 1993. Mobili-
zation, Participation, and Democracy in America. New York:
Macmillan.
Sharp, Clifford. 1981. The Economics of Time. Oxford: Martin
Robinson.
Silver, Brian D., Paul R. Abramson, and Barbara A. Ander-
son. 1986. "The Presence of Others and Overreporting of
Voting in American National Elections." Public Opinion
Quarterly 50:228-39.
Silver, Brian D., Barbara A. Anderson, and Paul R. Abram-
son. 1986. "Who Overreports Voting?" American Political
Science Review 80:613-24.
Sorauf, Frank J. 1988. Money in American Elections. Glenview,
IL: Scott, Foresman and Little Brown.
Stigler, George J., and Gary S. Becker. 1977. "De gustibus non
est disputandum." American Economic Review 67:76-90.
Strate, John M., Charles J. Parrish, Charles D. Elder, and Coit
Ford III. 1989. "Life Span Civic Development and Voting
Participation." American Political Science Review 83:445-63.
Teixeira, Ruy A. 1992. The Disappearing American Voter. Wash-
ington: Brookings Institution.
Thorndike, Robert L., and George H. Gallup. 1944. "Verbal
Intelligence of the American Adult." Journal of General
Psychology 30:75-85.
Verba, Sidney, and Norman H. Nie. 1972. Participation in
America. New York: Harper & Row.
Verba, Sidney, Norman H. Nie, and Jae-on Kim. 1978. Partic-
ipation and Political Equality: A Seven Nation Comparison.
Cambridge: Cambridge University Press.
Verba, Sidney, Kay Lehman Schlozman, and Henry E. Brady.
1995. Voice and Equality: Civic Voluntarism in American Politcs.
Cambridge: Harvard University Press.
Verba, Sidney, Kay L. Schlozman, Henry E. Brady, and
Norman Nie. 1993. "Citizen Activity: Who Participates?
What Do They Say?" American Political Science Review 87:
303-18.
Volgy, Thomas J., and John E. Schwarz, 1984. "Misreporting
and Vicarious Political Participation at the Local Level."
Public Opinion Quarterly 48:757-65.
Wald, Kenneth D. 1992. Religion and Politics in the United
States. 2d ed.; Washington: CQ Press.
Weiss, Carol H. 1968. "Validity of Welfare Mothers' Interview
Responses." Public Opinion Quarterly 32:622-33.
Wolfinger, Raymond E., and Steven J. Rosenstone. 1980. Who
Votes? New Haven: Yale University Press.
Wright, Erik Olin. 1985. Classes. London: Verso.

Henry E. Brady is Professor of Political Science and Public Policy, University of

California at Berkeley, Berkeley, CA 94720.

Sidney Verba is Professor of Government, Harvard University, Cambridge, MA

02138.

Kay Lehman Schlozman is Professor of Political Science, Boston College, Chestnut

Hill, MA 02167.

294

This content downloaded from 200.130.19.155 on Thu, 18 Jan 2018 21:27:20 UTC
All use subject to http://about.jstor.org/terms